Eli Lilly & Co. v. Zenith Goldline Pharmaceuticals, Inc.

FINDINGS OF FACT AND CONCLUSIONS OF LAW

YOUNG, District Judge.

TABLE OF CONTENTS

FINDINGS OF FACT

I.The Parties.830

II.Background in the Relevant Field Prior to the Prosecution of the ’382 Patent.831

A. Schizophrenia.831

B. Early Drug Treatment — Typical Antipsychotic Drugs.831

C. Clozapine: The First Atypical Antipsychotic.832

D. The Search for a Safe, Atypical Antipsychotic Drug.832

E. Lilly’s Attempts to Discover a Safe, Atypical Antipsychotic Drug.832

F. The Discovery of Olanzapine .834

III. Prosecution History of the ’382 Patent.835

A. The ’143 Patent Application .835

B. The ’348 Continuation Application.837

IV. The Validity of the ’382 Patent.841

A. Anticipation.841

1. Anticipation by

Chakrabarti

1980a.841

2. Anticipation by

Schauzu

.843

B. Obviousness.844

1. The Scope and Content of the Prior Art.844 a. Clozapine and Clozapine-Like Molecules .844 b. The’574 Patent .844

c. The

Chakrabarti

Articles.845

(1)

Chakrabarti 1980a

.845

(2)

Chakrabarti

1982.845

(3)

Chakrabarti

1989.846

d. The

Sullivan and Franklin

Article .846

2. Ordinary Skill in the Art.846

3. The Differences Between the Claimed Invention and the Prior Art.847

*826

4. Motivation Provided by the Prior Art to Make Olanzapine.847 a. Compound ’222 as the Beginning Compound.847

(1) Compound ’222’s Activity.847

(2) Hydrogen as a Preferred Substituent.848

(a) Compound 6 .848

(b) Compound 36 .848

(c) Compound 34 .849

b. Changing the 2-ethyl in Compound ’222 to a 2-methyl.849

c. Replacing the Fluorine Atom in Flumezapine with a Hydrogen Atom.850

5. Reasonable Expectation of Success.850

6. Composition and Method Claims .851

7. Secondary Considerations. 00 or to

a. Long-Felt Need. 00 cn co

b. Failure of Others. 00 cn co

c. Commercial Success . 00 cn co

d. Industry Acclaim. 00 en co

e. Unexpected Differences Between Compound ’222 and Olanzapine — The Dog Studies.

Principles Toxicology.

(2) Lilly’s and Defendants’ Dog Toxicology Studies ....

(a) Lilly’s D07290 Dog Study.

(b) Zenith’s Dog Study — The MPI Study.

(c) DRL’s Dog Study — The Calvert Study.

(3) Criticisms of Lilly’s Dog Study Evidence.

(a) The Dog as a Model..

(b) Total Cholesterol as a Tested Parameter.

(e) Randomization.

(d) The Length of the Study.

(e) The Number of Dogs and Analysis by Sex.

(f) The Dosage Used.

(g) The Use of Equal Doses of Olanzapine and Compound ’222 . O

<0

00

(h) How the Study Was Conducted. O

ZD

00

i) Good Laboratory Practices Were Followed O

ZD

00

ii) Double Rations . O

ZD

00

(i) The Results of Lilly’s and Zenith’s Studies. 1-1

SO

00

i) Statistics Experts for All of the Parties Found a Statistically Significant Cholesterol Increase in the Compound ’222-Treated Dogs. 00

ii) The Repeated Measures Analysis Was Appropriate. 00

iii) Dunnett’s Test Does Not Show that the Effect of Compound ’222 Is Small.

iv) Dr. Gibbons’ UPL Test Does Not Show that the Effect of Compound ’222 Is Small .

(j) The Cholesterol Results.

i) The Increase in Cholesterol Is Biologically Significant. 00

ii) The Increase in Cholesterol Was Not Caused by Other Factors. 00

a) The Female Estrous Cycle. 00

b) Double Rations . 00

c) The Alleged Hypothyroid Dog. 00

iii) The Reference Range. 00

iv) Other Statistically Significant Changes. 00

v) Dog Data Excluded from the Study Did Not Effect the Study Findings. 00

f. Prolactin' as a Previously Unconsidered Unexpected Result 00

(1) The Results of the MPI Study. 00

00 00

*827

[[Image here]]

(3) Olanzapine’s Effect on Prolactin. o t— oo

g. Unexpected Differences Between Olanzapine and Flumezapine. o 00

(1) Liver and Muscle Enzyme Test Results from the Flume-zapine Clinical Trials . t-00

(2) Liver and Muscle Enzyme Test Results from Olanzapine Clinical Trials. oo -3 > —

1

(3) Dr. Diamond’s Opinion. oo -q CO

C. Double Patenting. 00 «<3 co

D. Public Use. 00 —-3 oo

E. Inequitable Conduct. OO —q cr

1. Lilly’s Statements to the PTO Did Not Contradict Prior Statements It Made to the Swedish Board of Health. 00 —3 or

2. Lilly Did Not Believe that It Was Necessary to Compare Olanza-pine and Compound ’222 in a Second Species.

3. Dr. McGrath Believed the Effect of Compound ’222 on Cholesterol Was Significant Before Dr. Symanowski Performed a Statistical Analysis. c— oo

4. Dr. McGrath’s Draft Definition of “Clinical Pathological Significance” Was Not Lilly’s Standard for Clinical Significance. CO 00

5. Dr. Emmerson Did Not Mislead the Patent Examiner Regarding the Significance of the Results of the D07290 Study. 00 -q co

6. Lilly Did Not Mislead the PTO Regarding the Relation Between the D07290 Study in Dogs and the Potential Effects of Cholesterol in Humans. 00 00

a. Lilly’s Representations in the Response After Final. 00 00

b. Dr. Scruby’s Declaration . 00 CO

c. Dr. Means’ Declaration. 00 00

d. Dr. Tye’s Declaration. 00 00

7. Lilly Did Not Mislead the PTO by Withholding Individual Dog Data. CO 00 CR

8. Nonprotocol Bleed Data from Moribund Dog 240712 and from Dog 240692 After the End of the Study Taken in Connection with a Bone Marrow Biopsy Are Not Material. oo 00 05

9. A Statement in the Response After Final That There Was No Cholesterol Elevation in “Any” of the Olanzapine-Treated Dogs Was Not Intentionally Misleading. 00 00 -q

10. Individual Data from the Control Dogs Does Not Contradict Lilly’s Arguments to the PTO. 00 00 00

11. Individual Data from Two Dogs Offered Double Rations Did Not Confound the Study Results. 00 00 CO

12. The Change in the Statistical Analysis Protocol from Dunnett’s Test to Repeated Measures Was Proper. 00 00 CO

13. Dr. Symanowski Did Not Withhold Results Inconsistent with the Results He Presented to the PTO . f — i 05 00

*828

14. There Was Nothing Misleading About the Use of the Standard Error in Figure 3 of Dr. Symanowski’s Declaration. 00 CO to

15. Other Parameters Did Not Show Olanzapine to Be More Toxic than Compound ’222 . 00 CO CO

16. The Reference Range for Cholesterol Is Not Material and Was Not Intentionally Withheld from the PTO . CO 05 00

17. There Was No Inequitable Conduct Regarding the Failure to Disclose the Phase I Clinical Trials. 05 00

18. There Was No Inequitable Conduct Regarding Lilly’s Failure to Disclose the ’574 Patent and the

Chakrabarti

Articles. 00

a.

Chakrabarti 1980a

and the ’574 Patent. ^ 00

b.

Chakrabarti 1982

and

Chakrabarti 1989

. LO 00

CONCLUSIONS OF LAW

I. Controlling Authority. AO 0*5 00

A. Jurisdiction. ÍO 05 00

B. Federal Circuit Law Applies. CO 05 00

C. The Presumption of Validity. CO -05 00

D. Selection Inventions. 05 CO

II.- The Validity of the ’382 Patent. 1> 05 00

A. Anticipation — Lack of Novelty under 35 U.S.C. § 102 . t> 05 00

1.

ChakrabaHi 1980a

Does Not Anticipate the Claims of the ’382 Patent. 00 05 00

a.

Chakrabarti 1980a

Does Not Disclose a Genus . 05 05 00

b. The Preferences Expressed in

Chakrabarti 1980a

Do Not Lead a Person of Ordinary Skill in the Art to Envision Olanzapine. CD o o

2.

Schauzu

Does Not Anticipate the Claims of the ’382 Patent... CD O o

B. Obviousness. rH O 05

1. The Prima Facie Case . CO O 05

a. It Would Not Have Been Obvious to Start with Compound ’222 . rtf O 05

b. It Would Not Have Been Obvious to Modify Compound ’222 to Arrive at Olanzapine. ^ O 05

c. It Would Not Have Been Obvious to Modify Flumezapine . ^ O 05

d. There Was No Reasonable Expectation of Success. LO O 05

2. Objective Evidence of Nonobviousness — The Secondary Considerations. LO

a. Long-Felt Need. CO

b. Failure of Others. CD

c. Commercial Success . CD

d. Industry Acclaim. t>

e. Unexpected Results. C>

3. Conclusion Regarding Obviousness. 05 o

C. Double Patenting. 1. Domination. 05 0 O rH

2. The Double Patenting Issue .

D. Public Use.

*829

1. The HGAA, HGAB, and HGAC Phase I Clinical Trials Were Not a Public Use of Olanzapine .912

2. The HGAA, HGAB, and HGAC Phase I Clinical Trials Were Not a “Use” of the Claimed Invention.913

3. The HGAA, HGAB, and HGAC Phase I Clinical Trials Were “Experimental Uses” of Olanzapine.913

E. Inequitable Conduct rH 05

1.Materiality . rH 05

2.Intent to Deceive.•. 915

3. Lilly Did Not Commit Inequitable Conduct During the Prosecution of the ’382 Patent. CD rH 05

a. Lilly’s Nondisclosure of Prior Art Does Not Constitute Inequitable Conduct. CD i — 1 05

b. Dr. McGrath Believed the Cholesterol Results of the D07290 Study Were Significant. tH 05

c. Dr. Symanowski Did Not Conceal or Misrepresent the D07290 Study Data. 00 r — t 05

(1) The Use of the Repeated Measures Analysis. 00 i — I 05

(2) The Failure to Include Individual Dog Data. co rH 05

(3) The Failure to Include Nonprotocol Bleed Data. 00 tH 05

(4) The Failure to Inform the PTO of the Results of the Robustness Cheek. CD CD

(5) Figure 3 of Dr. Symanowski’s Declaration. CD CD

(6) The Failure to Disclose the Results of Other Parameters CD CD

d. Statements Made by Other Lilly Representatives Do Not Constitute Inequitable Conduct. CD

(1) Lilly’s Response to the Swedish Board of Health .... CD

(2) Dr. Emmerson’s Statement to the Examiner. CD

(3) Lilly’s Representations in the Response After Final.. CD

(4) Dr. Scruby’s Declaration. CD

(5) Dr. Means’ and Dr. Tye’s Declarations. CD

(6) Lilly’s Other Failures to Disclose Were Not Material CD

e. Conclusion Regarding Inequitable Conduct. CD

III. Summary of Conclusions. .922

Plaintiffs, Eli Lilly and Company and Lilly Industries Ltd., filed suit against the Defendants, Zenith Goldline Pharmaceuticals, Inc., Dr. Reddy’s Laboratories, Ltd., and Teva Pharmaceuticals USA, Inc. (collectively “Defendants”), for infringement of United States Patent No. 5,229,382 (“’382 patent”). The parties tried this case before the court from January 26, 2004, through February 12, 2004. Following the trial, the parties filed proposed findings of fact and conclusions of law, and responses thereto. The parties also filed post-trial briefs, which the court found helpful given the breadth and complexity of the disputed issues. The majority of the relevant briefing was submitted by May 12, 2004, with final submissions filed in February 2005.

Being duly advised, the court finds that Defendants have failed to prove by clear and convincing evidence that the ’382 patent is invalid, as anticipated under 35 U.S.C. § 102 , as obvious under 35 U.S.C. § 103 , under the doctrine of double patenting, or as barred by prior public use under 35 U.S.C. § 102 . The court further finds

*830

that Defendants have failed to prove by clear and convincing evidence that the ’382 patent is unenforceable due to inequitable conduct.

The court now issues its findings of fact and conclusions of law pursuant to Federal Rule of Civil Procedure 52(a):

FINDINGS OF FACT

1

I.The Parties

1. Eli Lilly and Company is an Indiana corporation engaged in the business of research, development, manufacture, and sale of pharmaceutical products throughout the world.

2. Lilly Industries Ltd., located in England, is a subsidiary of Eli Lilly and Company (the Plaintiffs are hereinafter collectively and individually “Lilly”).

3. Zenith Goldline Pharmaceuticals, Inc. (“Zenith”) is a Florida corporation having its corporate offices and principal place of business at 4400 Biscayne Boulevard, Miami, Florida 33137. Zenith’s Amended Answer to Complaint for Patent Infringement, Affirmative Defenses and Counterclaims, filed August 7, 2002, ¶ 3.

4. Dr. Reddy’s Laboratories, Ltd. (“DRL”) is a public limited liability corporation having its principal place of business at 7-1-27 Ameerpet, Hyderabad 500 016, India. DRL’s Amended Answer and Counterclaim, filed July 29, 2002, ¶ 3.

5. Teva Pharmaceuticals USA, Inc. (“Teva”) is a Delaware corporation with its principal place of business at 650 Cathill Road, Sellersville, Pennsylvania. Answer of Defendant Teva Pharmaceuticals USA, Inc., filed on March 18, 2003, ¶ 3.

6. On July 20, 1993, the United States Patent and Trademark Office (“PTO”) issued the ’382 patent which is entitled “2-Methyl-Thieno-Benzodiazepine.”

The ’382 patent was assigned to, and is owned by, Lilly. TX 1000; TX 1360.

7. The ’382 patent claims,

inter alia,

the chemical compound known as olanza-pine and methods of using olanzapine to treat schizophrenia. TX 1000, col. 12, claims 1, 2, 3, 7, 8, and 15.

8. Olanzapine, sold by Lilly under the trademark ZYPREXA® (“Zyprexa”), was approved by the United States Food and Drug Administration (“FDA”) in late 1996. Paul Tr. 139:16-19.

9. Zenith, DRL, and Teva filed Abbreviated New Drug Applications (“ANDAs”) under the Drug Price Competition and Patent Term Restoration Act of 1984, 98 Stat. 1585 (popularly known as the Hatch-Waxman Act), seeking approval to market generic copies of Lilly’s olanzapine products prior to the expiration of the ’382 patent.

10. Pursuant to 21 U.S.C. § 355 (j)(2)(B)(ii), Zenith, DRL, and Teva sent letters to Lilly to notify it that they had filed ANDAs for olanzapine in various dosages.

11. Lilly filed suit against Zenith, DRL, and Teva alleging infringement of the ’382 patent under 35 U.S.C. § 271 (e)(2)(A). The court consolidated Lilly’s suits against Zenith, DRL, and Teva into this single action.

See e.g.,

Entries of February 15, 2002; April 11, 2002; November 1, 2002; and March 14, 2003.

*831

12. Lilly seeks an order (1) prohibiting FDA approval of the Defendants’ generic olanzapine products prior to the expiration of the ’382 patent, in accordance with 35 U.S.C. § 271 (e)(4)(A); and (2) enjoining the Defendants from the commercial manufacture, use, offer to sell, sale, or importation of their olanzapine products, in accordance with 35 U.S.C. § 271 (e)(4)(B). Complaint, filed April 2, 2001, Prayer for Relief.

13. After the commencement of the suit, Zenith and DRL stipulated that “if the Court finds the ’382 patent valid and enforceable, then their actions constitute infringement.” Entry, December 2, 2003 at 2. The parties further stipulated that the only method-of-use claims to be tried in this case are claims 7 and 8 of the ’382 patent relating to the treatment of schizophrenia. Stipulation and Order entered on July 31, 2003 at p. 3. Therefore, the issues before the court are the validity and enforceability of claims 1, 2, 3, 7, 8, and 15 of the ’382 patent.

14. Teva did not participate in the trial but agreed to be bound by the decision of the court herein. Entry on Joint Stipulation and Staying Actions, July 16, 2003.

II. Background in the Relevant Field Prior to the Prosecution of the ’382 Patent

A. Schizophrenia

15. Schizophrenia is a chronic, debilitating mental illness that appears during late adolescence or early adulthood and essentially lasts the lifetime of the patient. Paul Tr. 109:17-110:1.

16. Some of the symptoms of schizophrenia include, but are not limited to, “positive” symptoms and “negative” symptoms. “Positive” symptoms include hallucinations, delusions, and thought disorders.

See, e.g.,

video clip at TX 1446.1. “Negative symptoms” include loss of emotional and mental functioning, loss of motivation, loss of normal emotional response to other people, slowness of thinking, memory deficits, changes in speech (speaking in a dull monotone), and difficulties with cognition, sustained attention, decision-making, and mental flexibility. Paul Tr. 108:2-25, 109:1-16; Schulz Tr. 2972:1-5.

B. Early Drug Treatment — Typical Antipsychotic Drugs

17. Prior to the discovery of antipsy-chotic medications in the 1950s, most schizophrenic patients were isolated from society and kept in large asylums. Paul Tr. 110:10-19. This pattern of lifetime institutionalization for ' schizophrenic patients began to change in 1952 with the introduction of chloropromazine and continued through the later introduction of haloperidol in the 1960s. TX 1398 at 746; Paul Tr. 111:2-9. These early antipsy-chotic medications (known as “typical” an-tipsychotics) showed substantial reduction of positive symptoms and allowed a number of patients to leave institutional settings. TX 1398 at 746; Paul Tr. 111:12-112:4.

18. However, the typical antipsychotic medications did little to treat the symptoms of schizophrenia and also induced a number of “sometimes severe and intolerable neurological side effects.” TX 1398 at 746; Paul Tr. 112:8-13. Such side effects associated with so-called extrapyramidal symptoms or “EPS” led to gross movement disorders, such as disfiguring tremors, stiffness, tics, and writhing. A particularly severe form of EPS, called “tardive dyskinesia,” persisted even after medications were withdrawn. Paul Tr. 113:22-114:1. In addition, patients experienced elevation of the hormone prolactin that led to breast engorgement and milk production in both male and female patients. Paul Tr. 112:5-115:12.

*832

C. Clozapine: The First Atypical An-tipsychotic

19. In the late 1960s and early 1970s, it became apparent that a drug called “cloza-pine” could treat the psychotic symptoms without EPS or prolactin elevation, causing it to be recognized as the first “atypical” antipsychotic. Paul Tr. 116:6-117:2; Nichols Tr. 2743:9-13.

20. In 1975, clozapine was withdrawn from the market in many countries because it was found to cause an often fatal blood disorder called “agranulocytosis” in approximately one percent of patients. This effect, and resulting withdrawal of the drug from the market, prompted people in the pharmaceutical industry to start looking for a drug like clozapine, but without the same side effect profile. Nichols Tr. 2743:14-2744:9; Paul Tr. 131:8-11; Schulz Tr. 3004:15-20. ,

D. The Search for a Safe, Atypical Antipsychotic Drug

21. The general failure for many years to find a clozapine replacement was reflected in the scientific literature. TX 1356, Nichols Tr. 2750:17-2752:2. The literature contains many reports of promising compounds which failed either for lack of efficacy or because -of toxic side effects. TX 1356; TX 1383; TX 1593; TX 1594; TX 1595; Nichols Tr. 2751:5-2769:24;

see also

Tupper Tr. 242:22-243:10, 246:9-247:8; TX 1595 (summarizing various scientific publications that reported the efforts of major pharmaceutical companies to find safe and effective antipsychotic medications).

22. In late 1989, the FDA approved clozapine for limited use in the United States notwithstanding its adverse side effects, as a therapy of last resort to be used only with careful blood monitoring. Paul Tr. 117:25-118:13.

23. In 1991, the New England Journal of Medicine reported that, “[djespite the extensive developmental effort in- this area, no alternative to clozapine has been identified that has clinical antipsychotic efficacy and no extrapyramidal neurologic side effects, but has a low risk of inducing other important toxic effects (bone marrow suppression' or seizures).” TX 1398 at 747.

24. Clozapine and certain other “typical” antipsychotics on the market such as chlorpromazine and haloperidol, had a “neuroleptic substituent,” an electron withdrawing group believed to be important for antipsychotic activity. The most common neuroleptic substituent has a halogen atom. The term “halogen” refers to either a fluorine or a chlorine atom. Tupper Tr. 414:9-418:18, 487:13-488:1.

E.Lilly’s Attempts to Discover a Safe, Atypical Antipsychotic Drug

25. In the late' 1970s and early 1980s, Lilly produced a novel class of compounds called thienobenzodiazepines. Tupper Tr. 428:23-429:25. Thienobenzodiazepines are tricyclic compounds having “thieno,” “ben-zo,” and “diazepine” rings fused together.

26. In 1975, Lilly filed a patent application related to these compounds. In 1978, the PTO issued two patents, U.S. Patent No. 4,115,568 (“the ’568 patent”), TX 1408, and U.S. Patent No. 4,115,574 (“the ’574 patent”), TX 3129, having identical technical disclosures, but different claims. Tup-per Tr. 429:16-20, 431:4-19. Dr. Jibán Chakrabarti (“Dr. Chakrabarti”) and Dr. David Tupper (“Dr. Tupper”) are listed as the named inventors on the ’568 patent and the ’574 patent. TX 1408; TX 3129.

27. The ’568 and ’574 patents described the members of this class of new compounds as “useful in the treatment of ... certain kinds of psychotic conditions ...” TX 3129, col. 13, 11. 62-66; TX 1408, col. 14,11. 41-42.

*833

28. The ’574 patent identifies the characteristics of the most preferred class of compounds within this family. TX 3129, col. 4, 11. 27-29. From this most preferred class, the patent identifies one particularly active compound as 2-ethyl-7-fluoro-10-(4’-methyl-l’-piperazinyl)-4H-thieno[2,3-b][l,5]benzodiazepine, known throughout this trial as “ethyl flumeza-pine.” Tupper Tr. 431:20-432:8.

29. Lilly, like many others in the field, modeled its lead candidates after cloza-pine, having a halogen atom in the molecule. Nichols Tr. 2751:5-2757:14; TX 1356 at 806; Tupper Tr. 413:21-414:9, 503:20-21; Pullar Tr. 200:20-201:3. Clozapine has a chlorine atom (Cl) at a position analogous to the fluorine atom (F) in ethyl flumezapine.

30. Other than the broad genus claimed in claim 1, Lilly’s ’574 patent claimed only compounds with a halogen. TX 3129, col. 39, claims 2-7; Tupper Tr. 433:5-14; Nichols Tr. 2749:5-9.

31. Beginning in the fall of 1974, Lilly made a compound like ethyl flumezapine, but without the fluorine atom, which corresponds to ethyl olanzapine, a compound otherwise known as “compound ’222”. Tup-per Tr. 426:23-427:10; TX 1205.

32. From the very earliest tests of the ethyl flumezapine and compound ’222 molecules, it was apparent that the fluorine-containing molecule (ethyl flumezapine) was much more active in tests believed to be relevant to potential antipsychotic activity than the molecule without the fluorine (compound ’222). Tupper Tr. 428:9-15.

33. After a year and a half of preclinical development work aimed at taking ethyl flumezapine into human clinical trials, disaster struck. In a six-month toxicology study with dogs, conducted at 4, 8, and 12 mg/kg using three dogs of each sex at each dose level, blood disorders, reminiscent of the potentially fatal blood disorder seen with clozapine in humans, were seen in dogs at all dose levels. TX 1035; TX 3421; Pullar Tr. 175:14-176:23. In particular, the toxicology tests in dogs showed widespread neutropenias in all dose groups and one anemia. Neutropenia is a reduction of white blood cells. Some of the dogs had reductions as much as 75% of their normal value. TX 3421; Emmerson Tr. 542:6-21.

34. In an effort to find a compound in the class free of this problem, Lilly conducted a comparative toxicology test between ethyl flumezapine and a closely related compound called “flumezapine.” TX 1003; Pullar Tr. 177:23-178:14. The difference between ethyl flumezapine and flumezapine is that in the two position on the thiophene ring where ethyl flumeza-pine has an ethyl group (-CH2-CH3), flumezapine has a methyl group (-CH3). Pullar Tr. 178:1-7.

35. During the comparative dog study, two of the ethyl flumezapine-treated dogs, but none of the flumezapine-treated dogs, developed blood problems. Pullar Tr. 180:11-19; TX 1003, 1004. As a result, Lilly terminated the development of ethyl flumezapine and commenced the development of flumezapine in early 1978 under the guidance of Dr. Ian Pullar (“Dr. Pul-lar”). Pullar Tr. 180:23-181:1; Hotten Dep. 71:12-72:10.

36. Lilly spent the next four years developing flumezapine through preclinical testing and initial safety testing in normal human volunteers. Pullar Tr. 185:20-188:12; TX 1008; TX 1010. Among other tests, Lilly conducted a six-month dog study of flumezapine, Emmerson Tr. 544:23-545:24; TX 1005, and a safety trial in normal human volunteers. TX 1010 at ZYP 177 1715-19. Then, in the spring of 1982, during the first trial of flumezapine in actual schizophrenic patients, administered at and below therapeutic doses, several patients experienced elevations in the muscle enzyme creatinine phosphokinase

*834

(ór “CPK”) and in a variety of liver enzymes. Pullar Tr. 188:13-189:14, 195:15-198:16; TX 1015 at ZYP 177 1998-99; Hotten Dep. 70:14-71:11.

37. The findings were reported by telephone to the FDA. TX 3259. That afternoon, Dr. Paul Leber in the Division of Neuropharmacology Drug Products at the FDA, halted U.S. clinical testing when he “advised that patients be withdrawn from the drug as soon as possible.” TX 3260.

38. Similarly, the British regulatory authority “expressed surprise at the magnitude of the increases in liver enzymes and especially the levels of CPK,” noting that- they “had not experienced anything similar.” TX 1615. After Lilly informed the agency of these results, the agency withdrew the United Kingdom (“U.K.”) authorization for clinical testing, and Lilly terminated the flumezapine clinical trials. Pullar Tr. 190:20-194:11; TX 1042.

39. Neither the failure of the flumeza-pine clinical trials nor the reasons for it (muscle or liver enzyme elevations) were publicly reported. Reith Tr. 927:9-11; Nichols Tr. 2776:18-2777:3; TX 1356 at 809.

F. The Discovery of Olanzapine

40. After the failure of flumezapine in clinical trials, a team led by Dr. Tupper at Lilly created another group of compounds in the same series in an effort to find another compound that could be developed as a clozapine replacement and that would not meet the same fate in the clinic as flumezapine. Tupper Tr. 441:23-443:2, 446:2-447:17; TX 1229-40.

41. The cause of flumezapine’s toxicity Was unknown. Tupper Tr. 443:3-20. Speculation abounded inside Lilly regarding possible causes, including reactions involving the “piperazine nitrogen” and the fluorine atom.

The distal piperazine nitrogen is dem-ethylated and the nitrogen and adjacent carbon oxidised as with other N-methylpiperazinyl compounds. If this is the source of the toxicity it is unlikely that it can be reduced without losing the neuroleptic activity. The 7-fluorine is replaced by hydroxy and methylthio groups .... [wjhether the source of this toxicity is the reduction in glutathione levels, a metabolic intermediate or the methylthio metabolite itself is at present, unknown. The hydroxylation metabolic pathway could also be implicated.

TX 3657 at ZY 80 94.

42.' Compounds with ethyl groups (like ethyl flumezapine and ethyl olanzapine (compound ’222)) were not considered for further development because the ethyl group was believed by some at Lilly to produce agranulocytosis in dogs. TX 3657 at ZY 80 94; Tupper Tr. 443:6-444:20; Hotten Dep. 79:6-80:8, 81:11-19.

43. Several fluorinated and unfluorinat-ed alternatives were made and tested, including the compound known as olanzapine (then known simply as “LY170053”). Hot-ten Dep. '42:5-16, 43:20-44:8, 45:1-17. Olanzapine was first synthesized in the U.K. by Terrence Hotten (“Mr. Hotten”), a research chemist at Lilly, on April 29, 1982. Tupper Tr. 446:2-14; TX 1229.

44. Olanzapine differs from flumeza-pine by having a hydrogen atom (conventionally not shown on structural diagrams) where flumezapine has a fluorine atom. Pullar Tr. 199:16-24; Hotten Dep. at 42:21-43:6. A number of people doubted that olanzapine would work because it lacked the halogen atom then known to be important to the activity of clozapine and believed to be important for activity in this series of compounds as well. Pullar Tr. 200:20-201:3; Tupper Tr. 451:7-20. Indeed, olanzapine was believed, based on a variety of preclinical tests, to be only about half as potent as flumezapine. Pullar Tr. 201:10-18; Tye Dep. 71:21-72:21, 75:21-

*835

76:12; TX 3657 at ZY 80 100-101, 103 (comparing two compounds).

45. In 1983, Lilly began by testing olanzapine in dogs — first in a three-month study and later in a one-year study. Em-merson Tr. 546:23-547-6, 552:14-17. During each of the studies, one dog developed a blood problem. Emmerson Tr. 546:19-24, 552:18-553:7. After extensive testing, Lilly determined that the effect in dogs appeared to be an immune response and “idiosyncratic” in nature, meaning it occurred due to the unusual sensitivity in individual dogs. Emmerson Tr. 549:22-550:23, 554:18-555:15. Only then did Lilly determine that it might cautiously proceed with human trials of olanzapine. Emmer-son Tr. 550:14-551:8, 555:6-15.

46. In 1986 and 1987, Lilly conducted Phase 1 clinical trials of olanzapine in healthy human volunteers in Indianapolis, Indiana. Goldberg Tr. 307:9-21; TX 3741; TX 3742; TX 3744. At the conclusion of these trials, Lilly proceeded to the litmus test of olanzapine — a clinical trial to test the compound’s efficacy in actual schizophrenic patients. Goldberg Tr. 326:25-327:20. These clinical trials took place in the U.K. TX 1058; TX 1064.

47. Toward the end of 1989, the clinical trials were promising. In the first test of the drug in actual schizophrenic patients, olanzapine appeared to be a safe and effective, atypical antipsychotic drug having a more favorable side effect profile than typical antipsychotics in terms of EPS and not producing the blood disorders in patients. Goldberg Tr. 335:19-20, 344:14-22, 345:6-9; TX 1063; TX 1064 at ZYP 520 983-84.

48. On January 18, 1990, the olanza-pine project team reported the success of the clinical trial to Lilly’s Research Management Staff (“RMS”), and the RMS agreed to “product commitment.” TX 1063 at ZYP 449 1132. After the project team report, the project was referred to with words like “AAA priority” and the “Manhattan Project” (in reference to the scientific push to develop the atom bomb), TX 3532, and Lilly scientists made plans for expanded clinical trials on the compound. Goldberg Tr. 338:21-340:7; TX 1063 at ZY 449 1128.

49. In the fall of 1990, Lilly conducted another comparative dog toxicology study prior to filing a patent application with the PTO. The study was designated D07290, and is known throughout this litigation as the “D07290 Dog Study” or “D07290 Study.”

50. The purpose of the D07290 Study was to determine over the course of a chronic treatment period whether there was a difference in the toxicity profile between olanzapine and compound ’222. Symanowski Tr. 664:4-665:13, 668:24-669:4, 2077:9-18; TX 3439 at ZYP 187 713, Item 4.

51. At the conclusion of the D07290 Study, Lilly claimed that the results of the study showed that olanzapine was unexpectedly superior to compound ’222 in that olanzapine did not cause a significant elevation in average mean cholesterol versus compound ’222. TX 1001.1 at FH 17-18.

52. The D07290 Dog Study is at the heart of this case and is discussed at length in this opinion.

III. Prosecution History of the ’382 Patent

A. The T43 Patent Application

53. On April 23, 1991, Lilly filed U.S. Application Serial No. 690,143 (“the Y43 application”). TX 1000, col. 1,11. 4-5.

54. Lilly had previously filed a patent application in the U.K. on April 25, 1990, and perfected the priority date. Pursuant to 35 U.S.C. § 119 , Lilly is entitled to the April 25, 1990 date for purposes of determining the relevant scope and content of the prior, art. Killworth Tr. 769:15-770:16, TX 1000.

*836

55. As part of the 143 patent application, Lilly was required to fill out an Information Disclosure Statement (“IDS”) to inform the Patent Examiner (“Examiner”) of the most closely related prior art. In Lilly’s IDS, Charles Ashbrook (“Mr. Ash-brook”), Assistant General Patent Counsel at Lilly, represented that the most closely related prior art was the matter claimed in the ’568 patent. TX 1001.1 at FH 44-45.

56. Though the patent applicants did not cite the Examiner to either the ’574 patent or Chakrabarti et ah,

k-Piperazi-nyl-10H-thieno[2,3-b][l, 5Jbenzodiaze-pines as Potential Neuroleptics,

J. Med. Chem. 23:878-84 (1980)

(“Chakrabarti 1980a

”), the Examiner performed a search of the prior art, found those references, and cited them in the Office Action. ' TX 1001.1at FH 48-53.

57. In the 1991 IDS, Lilly told the Examiner that compound ’222, described in the ’568 patent, is the adjacent homolog to olanzapine, but that olanzapine is patentable over the disclosure in the ’568 patent because of the “surprising biological differences of the 2-methyl derivative (olanzapine) over the 2-ethyl derivative (compound ’222).” TX 1001.1 at FH 44.

58. Also in the 1991 IDS, the patent applicants told the Examiner that flumeza-pine, which is the 7-fluoro derivative of olanzapine, “caused significant adverse effects when administered to humans” and that olanzapine “has an unexpectedly superior therapeutic profile.” TX 1001.1 at 45.

59. In the ’143 application, the applicants represented that: “In dog toxicity studies with a closely analogous compound 2-ethyl-10-(4-methyl-l-piperazinyIMH-thieno[2,3-b][l,5] benzodiazepine [compound ’222], at a dosage of 8 mg/kg, it was observed that four out of eight dogs showed a significant rise in cholesterol levels, whereas the compound of the invention did not show any rise in cholesterol levels.” TX 1001.1 at FH 17-18. This text corresponds to col. 3, 11. 29-36 of the ’382 patent, TX 1000, col. 3, 11. 29-36, and is a reference to the D07290 Dog Study. Plaintiffs’ Reply to Amended Answer to Complaint for Patent Infringement, Affirmative Defenses, And Counterclaims of Zenith Goldline Pharmaceuticals, Inc. (filed September 20, 2002) (“Lilly’s Reply”), ¶ 19.

60. In the ’143 application, Lilly stated that seventeen patients received flumeza-pine before the clinical trial was terminated after consultation with the FDA because of an unacceptably high incidence of raised enzyme levels in the treated patients. Specifically, “creatinine phosphoki-nase (CPK) and the liver enzymes, serum glutamate oxalacetic transmaninase (SGOT) and serum glutamate pyruvate transaminase (SGPT), estimated in the blood samples from the patients, were substantially in excess of normal values, indicating the possibility of toxicity.” TX 1001.1at FH 15. With respect to olanza-pine, Lilly stated only that “there is a low incidence of only mild and transient elevation of liver enzymes in patients treated with therapeutic doses, and plasma levels of ... CPK are lower than with flumeza-pine, indicating a lower adverse effect on muscular tissue.” TX 1001.1 at FH 17.

61. The Examiner reviewed the claims for compliance with the enablement requirement of 35 U.S.C. § 112 , the definiteness requirements of 35 U.S.C. § 112 , for novelty under 35 U.S.C. § 102 , for nonob-viousness under 35 U.S.C. § 103 , and for obviousness-type double patenting. TX 1001.1at FH 49-51, 53; Killworth Tr. 778:10-24, 779:5-21, 781:13-19.

62. On November 25, 1991, the Examiner issued an Office Action with respect to the ’143 application rejecting all of the claims. TX 1001.1 at FH 48-53.

63. In evaluating novelty, the Examiner rejected Lilly’s claims under 35 U.S.C.

*837

§ 102(b) as being “anticipated” by the ’574 patent, citing a portion of the text of the ’574 patent that was shared by the ’568 patent cited by Mr. Ashbrook in his IDS. Killworth 778:20-779:3; TX 1001.1 at FH 50.

64. The Examiner rejected all of the claims under 35 U.S.C. § 103 as obvious over the ’574 patent in view of

Chakrabarti 1980a.

TX 1001.1 at FH 51-53. The Examiner stated that “[i]t would have been obvious to one with ordinary skill in the art to replace the 2-ethyl substituent on the homologous species [compound ’222] taught in [the ’574 patent] with the 2-methyl substituent in order to obtain the instant compound [olanzapine] because

[Chakrabarti 1980a

] specifically suggests to one with ordinary skill in the art that this type of substituent is preferably [sic ] to increase [central nervous system] activity.” TX 1001.1 at FH 51-52.

65. The Examiner considered the description of the unexpected cholesterol results in the application to be “insufficient” because “1) [n]o controls were run, 2) there is no evidence that such data is statistically significant and 3) that such data is necessarily showing a significant beneficial effect to the patient.” TX 1001.1 at FH 52; Killworth Tr. 780:10-17.'

66. Finally, the Examiner rejected Lilly’s claims under the judicially created doctrine of “obviousness-type double patenting” as being unpatentable over the claims of the ’574 patent in view of

Chakrabarti 1980a

for the same reasons that the claims were alleged to be obvious from the text of the ’574 patent in view of

Chakrabarti 1980a

under 35 U.S.C. § 103 . TX 1001.1 at FH 53; Killworth Tr. 781:13-782:12.

B. The ’348 Continuation Application

67. Lilly responded to the Office Action on May 22, 1992, by filing a file wrapper continuation application to extend the time period for response to the rejection. TX 1001.1 at FH 98-100; Killworth Tr. 782:16-783:13. This application was assigned the application serial number 890,-348 (the “ ’348 application”). TX 1001.1 at FH 98.

68. The ’348 application contained the same disclosure, including the claims, as the 143 application. In accordance with standard PTO procedure, the ’348 application was assigned to the same Examiner who had examined the 143 application. Killworth Tr. 782:16-783:25; TX 1001.1 at FH 98-100.

69. On September 11, 1992, the Examiner issued another Office Action rejecting the claims for the same reasons set forth in the prior Office Action in the 143 application. Killworth Tr. 784:1-22; TX 1001.1 at FH 102.

70. On December 10,1992, three representatives of Lilly conducted a personal interview with the Examiner in the ’348 application. The Lilly representatives were Macharri Yorndran-Jones (“Ms. Vorndran-Jones”) and Joseph Jones (“Mr. Jones”) from Lilly’s legal department, and Dr. James Emmerson (“Dr. Emmerson”), a Lilly toxicologist. TX 1001.1 at FH 109-10.

71. At the interview, Lilly presented eight declarations of Lilly employees to the Examiner extensively describing the results of comparative tests of olanzapine and compound ’222, including the D07290 Dog Study. TX 1001.1 at FH 125-54.

72. These declarations were submitted by Dr. Nicholas Tye (“Dr. Tye”), Dr. Pul-lar, Dr. Nicholas Moore (“Dr. Moore”), Dr. Jeffrey Means (“Dr. Means”), Dr. Emmer-son, Dr. David Wong (“Dr. Wong”), Dr. David Scruby (“Dr. Scruby”), and Dr. James Symanowski (“Dr. Symanowski”). TX 1001.1 at FH 125-27 (Tye), FH 128-30

*838

(Pullar), FH 131-34 (Moore), FH 135-39 (Means), FH 140-42 (Emmerson), FH 143-45 (Wong), FH 146-48 (Scruby), FH 149-54 (Symanowski). Dr. Tye, Dr. Pul-lar, and Dr. Moore are pharmacologists; Dr. Means is a pharmacologist and toxicologist; Dr. Emmerson, as stated earlier, is a toxicologist; Dr. Wong is a biochemist; Dr. Scruby is a Lilly physician; and Dr. Symanowski is a statistician.

73. All declarations were submitted under 37 C.F.R. § 1.132 and all declarants swore that the statements made therein were true. TX 1001.1 at FH 125-54.

74. Dr. Moore’s declaration analyzed the results of behavioral tests in which animals were administered olanzapine or compound ’222. TX 1001.1 at FH 131-34. Dr. Moore concluded that in the conditioned avoidance response (“CAR”) test, olanzapine was more active. TX 1001.1 at 132. He also concluded that out of the four tests he conducted — the apomorphine-induced climbing and hypothermia tests in mice, the CAR test in rats, the catalepsy-induction in rats, and the locomotor activity in rats — “both compounds have very similar dopamine antagonist activity

in

vivo,

2

” TX 1001.1 at 134.

75. Dr. Pullar presented the results of

in vitro

testing of olanzapine and compound ’222 using binding assay tests. The results were presented in two tables showing the respective IC60 values for each compound in each of the tests. No argument was made that the compounds could be differentiated based on the results of these tests. TX 1001.1 at FH 128-30.

76. Dr. Tye’s declaration discussed his examination of the data set out by Dr. Moore and Dr. Pullar and his views on the D07290 Dog Study carried out by Dr. Means and the results thereof. TX 1001.1 at FH 124-27. Dr. Tye concluded that: (1) the data of Dr. Moore and Dr. Pullar suggested “little difference in the properties of [olanzapine and compound ’222]”; (2) the data from the dog toxicology study “show[ed] significantly increased levels of cholesterol in the case of [compound ’222],” (3) raised cholesterol levels resulting from compound ’222 are a serious disadvantage because cholesterol is a factor in coronary heart disease in humans, and (4) olanza-pine “is clearly and significantly superior to [compound ’222] so far as it has been shown in the toxicity study by Dr. Means to lack any tendency to raise cholesterol levels.” TX 1001.1 at FH 126.

77. Dr. Wong’s declaration evaluated data obtained from radioligand and binding studies of olanzapine and compound ’222. TX 1001.1 at FH 143-45. In part, Dr. Wong concluded that “both olan-zapine and [compound] ’222 can produce functional responses at 5-HT10 and 5-HT2 receptors.” TX 1001.1 at FH 145.

78. Dr. Emmerson oversaw the Lilly D07290 Dog Study. He stated in his declaration that he believed that “olanzapine displayed an unexpected and significant superior toxicological benefit over compound ’222” with respect to the elevation of serum cholesterol in female dogs dosed with 8 mg/kg of compound ’222 compared with dogs similarly dosed with olanzapine and the control dogs. TX 1001.1 at FH 142. Dr. Emmerson believed that this toxicological benefit was a critical property to consider in the safety analysis of the two compounds and that he “would not recommend the clinical development of a compound which significantly increases the serum cholesterol levels in dogs when there is a compound with similar activity which does not affect serum cholesterol levels.” TX 1001.1 at FH 142.

*839

79. Dr. Means was the study director for Lilly’s D07290 Dog Study. His declaration summarized the design, conduct, and findings of the study. TX 1001.1 at FH 134-39.

a. Dr. Means stated that “[sjuprisingly serum cholesterol levels in females of the 8 mg/kg-’222 treatment group were significantly increased when compared to the cholesterol levels of females in either the 8 mg/kg-olan-zapine treatment group or the control group.” TX 1001.1 at FH 137.

b. Dr. Means also stated that “[n]o significant differences in serum cholesterol levels were detected among the olanzapine treatment and control groups.” TX 1001.1 at FH 137.

c. Dr. Means concluded that “the toxicity of olanzapine and compound ’222 in beagle dogs is similar in many respects,” but that “[a]n unexpected difference between olanzapine and compound ’222 was the significant increase of serum cholesterol concentrations with time in female dogs given 8 mg/kg/day of compound ’222 compared to the cholesterol values in female dogs given 8 mg/kg/day [of] olanzapine and compared to the cholesterol values in female control dogs.” TX 1001.1 at FH 139.

80. Dr. Scruby stated that he reviewed Dr. Means’ and Dr. Symanowski’s declarations, and that those declarations provided the basis for Dr. Scruby’s clinical statements concerning the dog toxicology studies. TX 1001.1 at FH 146-48.

a. In his declaration, Dr. Scruby discussed his knowledge of the risk associated with elevated total cholesterol in humans and stated that “any factor which leads to an increase in serum lipids [] can have a significant deleterious effect on [cornary artery disease and the progression of atherosclerosis].” TX 1001.1 at FH 147-48.

b. Dr. Scruby concluded that based on the findings of the study in which the serum cholesterol female dogs treated with compound ’222 after 60 days averaged nearly 260mg/dl and “[i]n view of the overwhelming evidence in the literature that serum cholesterol in excess of 240 mg/dl is a significant contributor to the genesis of atherosclerosis,” he believed that “the significant elevation of serum cholesterol observed in female dogs treated with the ’222 compound could provide a marked clinical difference in the pathogenesis of coronary artery disease.” TX 1001.1 at FH 148.

81.Dr. Symanowski’s declaration summarized the statistical tests used in the D07290 Study. TX 1001.1 at FH 149-54. He reported that cholesterol levels in 8 mg/kg compound ’222 female dogs were noticeably elevated after day 25 of treatment and that statistical analysis indicated that these levels increased significantly over time when compared to both the 8 mg/kg olanzapine female dogs and the control group, whereas none of the olanza-pine-treated groups were statistically significantly different from control. TX 1001.1 at FH 150-51. He presented three graphs illustrating mean cholesterol levels over time. Figure 1 presented the group mean cholesterol values throughout the study for all of the groups of female dogs. TX 1001.1 at FH 152. Figure 2 presented the same information for the groups of male dogs. TX 1001.1 at FH 153. The data presented by Dr. Symanowski showed no significant difference between olanza-pine and compound ’222 with respect to cholesterol in male or low-dose female dogs. Figure 3 illustrated the mean values plus and minus one standard error for

*840

the cholesterol levels in female dogs dosed with 8 mg/kg of compound ’222 and olanza-pine, as well as the female control group. TX 1001.1 at FH 154.

82. Following the interview, the Examiner prepared a short, handwritten summary of the interview indicating that the Lilly representatives and the Examiner discussed the declarations. TX 1001.1 at FH 109. The Examiner reported in his summary that while “the reviewed data looked sufficient pending final review and analysis of the complete set of data,” an agreement was not reached as to patenta-bility. TX 1001.1 at FH 109.

83. Following the interview, the patent applicants presented to the PTO a document entitled “Response After Final.” TX 1001.1 at FH 112-24.

84. In the Response After Final, the patent applicants argued that “olanzapine exhibits the significant beneficial property of preserving the natural balance of cholesterol 'synthesis in the treatment of schizophrenia and schizophreniform disorders.” TX 1001.1 at FH 119-20.

85. In the Response After Final, the patent applicants also argued that Lilly’s “probative evidence of olanzapine’s superiority in one or more properties is sufficient to overcome

the jprima facie

obviousness rejection.” TX 1001.1 at FH 120.

86. To overcome a prima facie obviousness rejection, a patent applicant can respond to the rejection or the applicant can attempt to overcome the rejection by filing declaratory evidence. Sofocleous Tr. 959:15-960:6; Killworth Tr. 789:4-16.

87. While Lilly did not acquiesce to the Examiner’s determination that the claims were prima facie obvious, it did not advance specific argument to challenge this finding. TX 1001.1 at FH 117, 124 (indicating Lilly’s intent not to admit prima facie obviousness); Vorndran-Jones Tr. 1611:3-7; Killworth Tr. 789:4-790:2; Sofo-cleous Tr. 789:15-22.

88. In the Response After Final, Lilly made several significant statements.

a. “Dr. Means’ Declaration demonstrates that the blood cholesterol levels of female dogs treated with the ’222 derivative were significantly elevated when compared to the blood cholesterol levels of dogs treated with the 2-methyl (olanza-pine) compound.” TX 1001.1 at FH 120-21.

b. “No significant increase in cholesterol levels was observed in either the olanzapine treated dogs or the control dogs.” TX 1001.1 at FH 120.

c. “[Vjalid controls were included in the study as substantiated by Dr. Means’, Dr. Emmerson’s, and Dr. Symanowski’s Declarations.” TX 1001.1 at FH 121.

. d. “Statistically significant elevated blood cholesterol levels occurred in the ’222 treated dogs, as demonstrated by Dr. Symanowski’s Declaration.” TX 1001.1 at 121.

e. “Dr. Tye declares that pharmacolog-ically there is little difference between olanzapine and ’222; however, he states that olanzapine is clearly and significantly superior to the ’222 compound based on olanzapine’s toxicological benefit.” TX 1001.1 at FH 122.

89. On December 17, 1992, the Examiner issued a Notice of Allowability of all of the pending claims of the ’382 patent, meaning that he had determined that the requirements for patentability had been met and that all previous rejections were withdrawn. TX 1001.1 at FH 156.

90. The Examiner noted that the Notice of Allowability was in response to “Amendment B [Lilly’s Response After Fi

*841

nal] and the declarations filed 12/10/92.” TX 1001.1 at FH 156; Killworth Tr. 792:16, 793:16.

91. On July 20, 1993, the PTO issued the ’382 patent. TX 1000; TX 1360. The named inventors are Dr. Chakrabarti, Dr. Tupper, and Mr. Hotten.

IV. The Validity of the ’382 Patent

A. Anticipation

1. Anticipation by

Chakrabarti 1980a

92. Defendants argue the claims of the ’382 patent are anticipated by

Chakra-barti 1980a.

One of the authors of that scientific article is Dr. Chakrabarti of Lilly. TX 3465.

93. The

Chakrabarti 1980a

publication describes several variations of the 4-piper-azinyl-10H-thieno[2,3-b][l,5]benzodiaze-pine family of compounds and examines specifically three areas on the structure of the family of molecules. TX 3465; Reith Tr. 820:14-821:10; 827:18-828:3.

94. All compounds .examined in

Chak-rabarti 1980a

had a common structural nucleus with different substitutions at three places: the piperazine ring, the benzene ring, and the thiophene ring. TX 3465; Reith Tr. 827:18-828:3. The authors labeled these three areas for substitution “R”, “Rj”, and “R2”. TX 3465.

95. Only particular substituents for each disclosed compound are listed in a table. TX 3465 at 880-82.

96. In total the authors examined forty-five specific compounds (as opposed to a genus of compounds) in the 4-piperazinyl-10H-thieno[2,3-b][l,5]benzodiazepine family and 14 analogous 5-piperazinyl-substi-tuted 4H-thieno[2,3-b][l,4]benzodiaze-pines, created “[t]o compare the activity.” TX 3465 at 879, 880-82. Significantly, olanzapine was not one of the compounds the authors examined. TX 3465; Reith Tr. 828:7-9.

97. To arrive at their conclusions about which compounds were preferred, the authors tested the compounds by using three different animal behavioral tests: the mouse hypothermia test, the CAR test, and the rat catalepsy (“CAT”) test. TX 3465; Reith Tr. 821:16-826:5. Of these, the tests of special interest were the CAR and the CAT test. Pullar Tr. 254:4-11.

98. The CAR test evaluates the inhibition of a behavioral response in rats. Reith Tr. 822:19-823:13. In this test, inhibition is expressed on a scale from zero to five where a score of zero' reflects minor inhibition, and a score of five represents severe inhibition. Reith Tr. 823:8-13.

99. The CAR test was the only measure of potential antipsychotic activity, and if the compound did not achieve a CAR score of three or four at a dose of less than 30 mg/kg, it was not considered active. Nichols Tr. 2768:16-2769:11.

100. A “good score” in the CAR test— a three or four — indicates a desirable blockade of dopamine receptors. Reith Tr. 823:22-24; Pullar Tr. 222:25-223:1. Clozapine, the benchmark compound, had a ■ CAR score of three. LaVoie Tr. 1568:11-15.

101. The CAT test evaluates cataleptic behavior in rats following administration of a compound. The scores in this test reflect group scores, with a group consisting of eight animals. The higher the score the more catalepsy observed in the group. Reith Tr. 824:8-825:16.

102. Generally, for purposes of determining whether a compound has the potential to be an effective atypical antipsy-chotic, a scientist would like to see a “separation of activity” — i.e., a good score on the CAR test at a dose that would not get a high score on the CAT test. Reith Tr. 826:6-23; Pullar Tr. 254:4-15.

*842

103. However, a separation between the CAR and CAT scores only came into play if the first condition — a good CAR score — was met. Moore Dep. 82:5-22, 83:9-84:6, 86:5-14. The separation of activity between the CAR and CAT scores was thought to be relevant to the potential absence of EPS at an active or potentially therapeutic dose. The separation, by itself, is not a measure of activity. Moore Dep. 83:9-84:6, 86:5-14.

104.

Chakrabarti 1980a

identified five specific compounds (9, 12, 17, 29, and 34) that were “found to be more potent than clozapine and show similar, if less marked, separation of activity in [the CAR and CAT] tests.” TX 3465 at 878, 883.

105. The five preferred compounds had a CAR score of three or better. LaVoie Tr. 1568:16-18.

106. With respect to these five preferred compounds, the authors expressed a preference for specific, complete compounds; they did not express a preference for a genus of compounds having any and all combinations of the individual substitu-ents on those molecules. Nichols Tr. 2779:2-8 (“[A] preferred compound would have a combination of different substitu-ents that would lead it to have optimal activity. Removing one or more of those substituents would destroy that preference. So the compound would be taken as a whole.”); LaVoie Tr. 1567:11-19.

107. Four of the five preferred compounds (9, 12, 17, and 29) have a fluorine at the 7-position of the benzene ring (compound 29 contains two fluorine atoms, one at the 7-position and one at the 8-posi-tion). Tupper Tr. 437:4-11; Nichols Tr. 2749:9-18, 2780:2-9; TX 3465 at 880-81. The fifth, compound 34, does not contain a fluorine on the benzene ring, but contains a hydroxyethyl group on the piperazine ring. Tupper Tr. 437:12-16; Nichols Tr. 2749:9-18, 2780:11-19; TX 3465 at 881. These two components were generally recognized at the time of the publication of

Chakrabarti 1980a

to enhance antipsychotic activity. LaVoie Tr. 1564:6-1565:4; Nichols Tr. 2785:18-2787:22; TX 3123 at 396-97. Olanzapine does not include either of these substituents. Reith Tr. 902:5-8; LaVoie Tr. 1564:16-18.

108. The authors did not list compound 6, the freebase of compound ’222, as a preferred compound. TX 3465. Compound 6 had a CAR score of two. TX 3465 at 880; LaVoie Tr. 1568:11-15.

109. In addition to the preferred compounds, the authors discussed the authors’ preferred substituents. Reith Tr. 897:6-12. For example, the authors expressed a preference for position 7 by stating that “[t]he substitution of the phenyl ring with a halogen atom (Cl, F) at position 7 enhanced the activity” and that the “7,8-difluoro compound (29) retained good activity.” TX 3465 at 879, col. 2. The text never states a preference for a hydrogen at position 7, as is required for olanzapine. In addition, the authors mentioned three specific substituent groups thought to be helpful: three R groups (methyl, hydrox-yethyl, and hydroxypropyl), three Rx substitutions (7-F, 7-C1, 7,8-di-F, explained above), and three R2 groups (2-methyl, 2-ethyl, and 2-isopropyl).

3

TX 3465 at 879. No possible combination of these preferred substituents, i.e., a preferred R with a preferred Rj and a preferred R2, would generate olanzapine because all of these combinations contain a fluorine (F) or a chlorine (Cl) atom at position 7 where olanzapine has only a hydrogen. Reith Tr.

*843

897:16-898:21; LaVoie Tr. 1566:2-12; Nichols Tr. 2777:4-2778:6; TX 3465 at 879.

110. Because none of the forty-five specific compounds disclosed in

Chakrabarti 1980a

include olanzapine, none of the five preferred compounds disclosed in

Chakra-barti 1980a

include olanzapine, and none of the preferred substituents include a hydrogen at position 7 as is required for olanza-pine, the court finds

Chakrabarti 1980a

does not describe olanzapine.

111. The court further finds that one of ordinary skill in the art, applying the preferences expressed in

Chakrabarti 1980a,

would not envision olanzapine.

112. The composition and method claims of the ’382 patent are directed to dosage forms containing specified amounts of olanzapine (e.g., claim 15) and methods of treating patients suffering from schizophrenia with specified doses of olanzapine (e.g., claim 8). Such dosage forms and methods are not described in

Chakrabarti 1980a

for any of the compounds disclosed in that article.

See

Findings of Fact ## 183-84.

2. Anticipation by

Schauzu

113. DRL also argues that olanzapine is described as compound 11 in a scientific article entitled Schauzu, H.G. and Mager, P.P.,

A Free-Wilson Study of k-Piperazi-nyl-1 OH-thienobenzodiazepine Analogues,

38 Die PhaRmazie 562 (1983)

(“Schauzu

”).

114. The biological data in

Schauzu

comes from Chakrabarti, J.K.,

et al., Effects of Conformationally Restricted lp-Pi-perazinylr-lOH-thienobenzodiazepine Neuroleptics on Central Dopaminergic and Cholinergic Systems,

J. Med. Chem. 1133(1982)

(“Chakrabarti 1982

”). Nichols Tr. 2793:20-2794:9; LaVoie Tr. 1506:9-18; Reith Tr. 902:9-20.

115. The compounds disclosed in

Chak-rabarti 1982

were fluorinated piperazine compounds. This means that the compounds had a fluorine in the 7-position and two nitrogens in the top ring (the pipera-zine ring). Nichols Tr. 2793:7-16.

116. The structure drawn in

Schauzu

is not a fluorinated piperazine compound because it is missing both a fluorine atom at the 7-position and one of the nitrogen atoms in the top ring. Nichols Tr. 2792:22-2795:19. Thus, the structure drawn in

Schauzu

is a piperidine compound- — not a piperazine as the title of the article otherwise sets forth. LaVoie Tr. 1532:5-19; Nichols Tr. 2789:3-23.

117. A piperidine compound has one nitrogen in the top ring (the piperidine ring), whereas a piperazine compound, as noted above, includes a second nitrogen substituent in the top ring (the piperazine ring). Nichols Tr. 2789:16-2790:1; LaVoie Tr. 1500:16-1501:10.

118.

Schauzu

was abstracted by both

Chemical Abstracts

and

Beilstein

as disclosing piperidine compounds. Nichols Tr. 2791:10-2792:18; LaVoie Tr. 1532:3-1535:17.

119. Olanzapine is a piperazine compound. Nichols Tr. 2795:12-16.

120. The structure drawn in

Schauzu

does not include olanzapine since the structure does not include a second nitrogen in the top ring. Nichols Tr. 2792:22-2795:19.

121. The biological data from

Chakra-barti 1982

does not include olanzapine because olanzapine is an

unflominated

pi-perazine compound. Nichols Tr. 2789:24-2790:1, 2795:12-16.

122. In order to find that olanzapine is described as compound 11 in

Schauzu,

one with ordinary skill in the art would have to mentally insert a nitrogen atom into the structure depicted in

Schauzu,

thereby converting it into a piperazine compound (in this case, olanzapine), yet ignore the fact that the biological data reported in the

Schauzu

article was from

Chakrabarti

*844

1982,

which discussed only fluorinated compounds.

Compare

LaVoie Tr. 1500:20-1501:2, 1504:4-1507:5,

with

Nichols Tr. 2793:20-2794:9, 2789:16-2790:1; Reith Tr. 902:9-20.

123. Olanzapine is not described as compound 11 in

Schauzu.

Nichols Tr. 2789:3-5.

B. Obviousness

124. Lilly contends the discovery of olanzapine and its unique properties represent a nonobvious selection invention within the broad genus of compounds disclosed in the ’574 and ’568 patents. In other words, the ’382 patent, which specifically claims olanzapine, is a “species” falling within the broad “genus” claimed in the ’574 patent. Defendants contend that the claims of the ’382 patent are obvious under 35 U.S.C. § 103 over the disclosure of olanzapine in the ’574 patent in combination with

Chakrabarti 1980a.

They also contend the claims of the ’382 patent are obvious over the disclosure of flumezapine in the ’574 patent in combination with

Chakrabarti 1980a

and the

Sullivan and Franklin

article discussed infra., Findings of Fact § TV.B.l.d.

1. The Scope and Content of the Pri- or Art

a. Clozapine and Clozapine-Like Molecules

125. The prior art included numerous failures to find a safe, atypical antipsychotic drug.

See

Findings of Fact § II.D.

126. As mentioned, clozapine, the first atypical antipsychotic, was withdrawn from the market in 1975. Thus, the challenge was to find a clozapine-like molecule that produced the benefits of clozapine without the adverse side effects. Nichols Tr. 2743:14-2744:9; Paul Tr. 131:8-11; Schulz Tr. 3004:15-20.

127. The prior art confirmed that small structural changes in clozapine-like molecules led to unpredictable changes in properties. Changing the position of the chlorine atom in clozapine changed it from an atypical antipsychotic to a typical antipsy-chotic. TX 3465 at 878 (“[Clozapine’s] 2-chloroisomer HF-2046 behaves like a classical neuroleptic ...”); Nichols Tr. 2796:16-2797:3; LaVoie Tr. 1543:23-1544:11; LD 45; LD 94. Changing the ring structure of clozapine led to variable and unpredictable toxicities. Tilozepine caused seizures. TX 1356, compound 3; TX 1365 at col. 12, claim 12; TX 3772 at 394 (tilozapine (NT 104-252) clinical trials terminated); Nichols Tr. 2752:15-2754:11; LaVoie Tr. 1547:11-15, 1549:24-1551:3, 1552:115-19. Fluperlapine caused agranu-locytosis in a some patients. TX 1356 at 809, compound 2; TX 1387 at 155; Nichols Tr. 2745:18-2747:9, 2751:13-2752:14; LD 87. Removal of the halogen atom from clozapine and fluperlapine (yielding perla-pine) destroyed antipsychotic activity. TX 1317 at 712, col. 2; TX 3124, compound lc; Tupper Tr. 420:1-14; Nichols Tr. 2744:22-2745:17, 2747:10-2748:10; LaVoie Tr. 1545:8-18; LD 142; LD 143.

b. The ’574 Patent

128. As noted, the ’574 patent, issued on September 19, 1978, describes a family

of

chemical compounds of thieno[l,5]benzo-diazepines that have useful central nervous system (“CNS”) activity. TX 3129 (abstract); Reith Tr. 848:10-15. The unique properties of these compounds render them particularly “useful in the treatment of mild anxiety states and certain kinds of psychotic conditions such as schizophrenia and acute mania.” TX 3129, col. 13:62-66; Reith Tr. 853:12-20; Nichols Tr. 2839:19-2840:1.

129. The text of the ’574 patent was written in 1975.

130. The first compound specifically mentioned in the ’574 patent is ethyl

*845

flumezapine, which is the sole compound identified as “particularly active.” The chemical name of that compound is 2-ethyl-7-fluoro-10-(4'-methyl-l'-piperazi-nyl)-4H-thieno[2,3-b][l,5]benzodiazepine. TX 3129, col. 4:30-34. Ethyl flumezapine has a fluorine atom on the benzene ring. This fluorinated compound is the only compound that the ’574 patent specification describes in actual examples of pharmaceutical formulations. TX 3129, Examples 37-40, col. 37, 1.62 — col. 39, 1.12; Tupper Tr. 431:20-432:22; Nichols Tr. 2749:5-9.

131. The specification identifies about one hundred other compounds as examples of the compounds of the invention, including flumezapine and compound ’222.

See

TX 3129, col. 4-6, 15-37. No biological data is reported, however, for any of the compounds and no specific compound, other than ethyl flumezapine, is identified as being particularly active. TX 3129.

132. While the ’574 patent claims un-halogenated compounds among the millions of compounds described in claim 1, the patent expresses a preference for halogen-containing compounds and specifically those with a halogenated substituent on the benzene ring in a location analogous to the chlorine in clozapine. TX 3129, col. 4:30-33, col. 39:42-62; Tupper Tr. 433:5-14; Nichols Tr. 2749:5-9.

133. In fact, the patent contains six claims defining specific compounds each containing a fluorine or a chlorine atom. Tupper Tr. 433:5-14; Nichols Tr. 2749:5-9; TX 3129, claims 2-7, col. 39:42-62.

134. Olanzapine is one of the millions of compounds within the scope of claim 1 of the ’574 patent. TX 3129, col. 39:15-40; Reith Tr. 850:3-851:2; Hotten Dep. 160:10-22.

c. The

Chakrabarti

Articles

135. The

Chakrabarti

publications also express a preference for halogen-containing compounds.

Chakrabarti 1980a,

previously discussed in Findings of Fact § IV.A.1, is also central to the obviousness inquiry.

(1)

Chakrabarti 1980a

136.

Chakrabarti 1980a

reports on the

in vivo

behavior and toxicity testing of some individual compounds within the generic family of the thienobenzodiazepines. TX 3465.

137. The authors expressed a preference for a halogen substituent at the 7-position of the phenyl ring. Tupper Tr. 434:12-435:3; TX 3465 at 879; Finding of Fact # 109.

138. The five preferred compounds identified by the authors (compounds 9, 12, 17, 29, and 34) have CAR scores equal to or better than clozapine’s score of three. Tupper Tr. 436:1-437:3; TX 3465 at 880-82. Compounds 9, 12, 17, and 29 have a fluorine at the 7-position, and compound 34 has a hydroxyethyl piperazine group on the piperazine ring. TX 3465; Nichols Tr. 2780:6-18; Findings of Fact ## 105, 107.

139. The authors also report that “[a] short alkyl substitution ([methyl, ethyl, isopropyl]) at position 2 of the thiophene ring seems to increase the activity.” TX 3465 at 879, col. 2. The authors compare the effect of a short alkyl substitution at position 2 versus a t-butyl substitution, a 2-hydrogen substitution, or other higher alkyl substitution. TX 3465 at 879, col. 2; Tupper Tr. 486:2-25. The only series of data that allows for each of these comparisons is based on results from fluorinated compounds. Tupper Tr. 435:10-25.

(2)

Chakrabarti 1982

140. In

Chakrabarti 1982

the authors further confirmed the belief that neuroleptic substituents, such as fluorine atoms, were necessary for good antipsychotic activity. Each of the twelve thienobenzo-diazepines described contains a fluorine

*846

atom at the 7-position. Reith Tr. 916:2-16; Nichols Tr. 2749:18-24; TX 3131 at 1135. The article specifically identifies fluorinated compounds “2 and 9” (flumeza-pine and ethyl flumezapine) as having “potent ... activity.” TX 3131 at 1137; Nichols Tr. 2749:18-24.

(3)

Chakrabarti 1989

141. In J.K. Chakrabarti et al.,

Synthesis and Pharmacological Evaluation of a Series of I-Piperazinylpyrazolo [3,4-b]- and-[I,3-b][l,5]benzodiazapines as Potential Anxiolytics,

J. Med. Chem. 2573 (1989)

(“Chakrabarti 1989”),

the authors reflected the ultimate preference for a fluorine substitution in this class of molecules by reporting that the fluorinated compound flumezapine “was chosen as a candidate for clinical trial.” Reith Tr. 922:25-923:3; Nichols Tr. 2749:25-2750:3; TX 3132 at 2574.

d. The

Sullivan and Franklin

Article

142. Sullivan and Franklin,

In Vitro Thiomethylation,

DRug, Metabolism, And Disposition 276 (1985)

(“Sullivan and Franklin

”), reports a study in which flumezapine was administered to dogs and rats. The study found that a methylthio metabolite arose from the metabolism of flumezapine in the test animals. Nichols Tr. 2774:20-2775:10, 2843:23-2844:3; TX 3161.

143. There is some toxicity associated with compounds that are metabolized via methylthio metabolites. TX 3161; Nichols Tr. 2844:8-13.

Sullivan and Franklin

does not teach that the methylthio metabolic product is toxic. Reith Tr. 928:13-16; Nichols Tr. 2775:12-18. In addition, the article does not specifically state that flumezapine is toxic. Reith Tr. 927:9-11; Nichols Tr. 2775:11-13.

144. The article describes that during the metabolism of flumezapine, the fluorine at the 7-position is replaced by a methylthio group. Thus, one with ordinary skill in the art may have been motivated, based

on

the resulting potential toxicity from the metabolic process as described in the article, to consider developing a compound without the fluorine atom. Reith Tr. 858:22-859:6. There is, however, nothing in the article to suggest that a hydrogen atom in place of the fluorine atom at the 7-position (yielding olanza-pine) would be desirable, Nichols Tr. 2776:5-11, or that to make such a substitution would avoid the formation of the methylthio metabolite. Reith Tr. 929:11-18.

145. The article does not teach that replacing the fluorine with a hydrogen would stop the formation of the methylthio metabolite. Indeed, acetaminophen (Tylenol®), a non-fluorinated compound, also forms a methylthio metabolite. Reith Tr. 928:17-928:21; Nichols Tr. 2776:5-11.

2. Ordinary Skill in the Art

146. A person of ordinary skill in the art in this case would be a scientist with a Ph.D. in medicinal chemistry, pharmacology, or a similar discipline. Reith Tr. 819:23-820:6; Nichols 2850:25-2851:4; La-Yoie 1527:22-1528:1.

147. Lilly’s expert, Dr. David Nichols (“Dr. Nichols”), is personally familiar with the search for a safe, atypical antipsychotic, the relevant scope and content of the prior art, and the capabilities of a person of ordinary skill in the art. The court qualified him as an expert in medicinal chemistry. Nichols Tr. 2737:13-14.

148. Dr. Nichols obtained a Ph.D. in medicinal chemistry from the University of Iowa in 1973. Nichols Tr. 2733:25-2734:2; TX 3764.1. Dr. Nichols worked in the 1980s on developing antipsychotic compounds with an atypical activity profile; that is, active antipsychotics, but with few side effects. Nichols Tr. 2735:13-2736:17.

*847

Dr. Nichols has taught the subject of anti-psychotics for many years, including during the time period of the prior art at issue in this case. Nichols Tr. 2734:8-21. Dr. Nichols also has personal experience with the behavioral tests and other screening tests relied upon in

Chakrabarti 1980a.

Nichols Tr. 2735:25-2736:7. The focus of Dr. Nichols’ research, however, has been hallucinogens and dopamine receptors. Nichols Tr. 2812:24-2813:6.

149. The court qualified Zenith’s expert, Dr. Maarten Reith (“Dr.Reith”), as an expert in the field of molecular pharmacology and neuropharmacology. Reith Tr. 815:25-816:16. Dr. Reith has extensive experience with respect to dopamine and serotonin receptors, TX 3170, the specific areas of the brain at which the effects of antipsychotics are observed. - TX 1000, col. 2:45-57; col. 7:6-59. Further, Dr. Reith demonstrated substantial knowledge of the tests used in

Chakrabarti 1980a,

as well as those reported in the declarations of Lilly’s scientists.

See e.g.,

Reith Tr. 821:16-827:17, 865:1-866:24.

3. The Differences Between the Claimed Invention and the Prior Art

150. Structurally, olanzapine differs from clozapine in that olanzapine has a methyl-substituted thiophene ring in place of the benzene ring in clozapine. Olanza-pine also has a hydrogen in place of the chlorine on its benzene ring.

See e.g.,

LD 52; LD 80. Clozapine caused agranulocy-tosis in humans. Paul Tr. 117:3-24.

151. Olanzapine differs structurally from ethyl flumezapine by replacement of the fluorine and ethyl group in ethyl flumezapine with a hydrogen and methyl group respectively.

See e.g.,

LD 48; LD 52. Ethyl flumezapine caused widespread blood problems in dogs; olanzapine did not. Nichols Tr. 2798:4-2800:25.

152. . Olanzapine differs structurally from flumezapine, described as ■ the lone clinical trial candidate out of this series of compounds, by substituting a hydrogen atom for the fluorine atom in-flumezapine at the 7-position of the benzene ring.

See e.g.,

LD 49; LD 52; Nichols Tr. 2801:1-2802:11.

153. Olanzapine differs structurally from its ethyl analog, compound ’222, by replacement of the ethyl group with a methyl group at the 2-position of the thio-phene ring.

See e.g.,

LD 50; LD 52. In Lilly’s D07290 Study, compound ’2212 caused a significant increase in cholesterol in female beagle dogs; olanzapine did not. TX 1001.1 at FH 149-54; McGrath Dep. 85:13-89:12.

4. Motivation Provided By the Prior Art to Make Olanzapine

a. Compound ’222 as the Beginning Compound

154. The ’574 patent specifically names over one hundred compounds, including compound ’222. TX 3129, col. 4-6, 15-37, Example 26(a). The patentees did not differentiate compound ’222 from the other exemplary compounds in the way that they did with ethyl flumezapine. TX 3129, col. 4:30-34.

155. Compound ’222 is not halogenated and, therefore, does not satisfy the preference expressed in the ’574 patent itself, in

Chakrabarti 1980a, Chakrabarti 1982,

or

Chakrabarti 1989.

Nichols Tr. 2748:11-2750:9, 1772:18-1773:2; Reith Tr. 898:12-899:10, 914:7-915:4, 916:2-16, 922:9-923:3; LaVoie Tr. 1566:5-12.

(1) Compound ’222’s Activity

156.

Chakrabarti 1980a

did not provide specific motivation to use compound ’222 as a starting point for further research. In fact, the article reported that com

*848

pound ’222, corresponding to compound 6 in Table I, did not have the minimum CAR score of three required for a suitable atypical antipsychotic agent and thus was not considered a preferred compound. Pullar Tr. 227:1-22; LaVoie Tr. 1568:11-1569:3; TX 3465 at 880-82.

(2) Hydrogen as a Preferred Substi-tuent

157.

Chakrabarti 1980a

also did not provide specific motivation to substitute a hydrogen in the 7-position. One skilled in the art would not have recognized from compounds 6, 34, and 36 in

Chakrabarti 1980a

that a hydrogen in the 7-position was a desirable substituent.

(a) Compound 6

158.

Chakrabarti 1980a

reported that compound 6 had a CAR score of two at 10 mg/kg. LaVoie Tr. 1568:11-1569:3; Pullar Tr. 227:1-22; Tupper Tr. 437:22-438:2; Nichols Tr. 2772:18-2773:2; TX 3465 at 880-81.

159. A better CAR score for compound 6 could not be achieved by increasing the dose, since higher doses would cause the test animal to completely stop responding to the test stimuli due to muscular incoor-dination. Pullar Tr. 223:8-224:2. Dr. Pul-lar explained that increasing the dose caused the test animals to completely stop responding to the test — a distinctly negative outcome. These results, based on tests with the compound administered orally, were corroborated with other results using intraperitoneal (IP) administration. Increasing the dose in both cases could not increase the activity of compound 6 above the substandard CAR score of two. Pullar Tr. 224:18-228:3; TX 3470.

160. The evidence does not support the argument that a person of ordinary skill in the art would presume that compound 6’s corresponding free base (compound ’222) would be more active. The difference between a maleate salt and a free base is just their physical forms, with salts typically being used for ease of handling. It is the same molecule in both forms. Tupper Tr. 504:20-505:8. As Dr. Tupper explained, the form (salt or free base) does not fundamentally change the pharmacological action. Tupper Tr. 505:5-8. There are some differences in the way the body processes the salt of a compound versus its free base due to the varying size of the particles. But if administered to achieve equal concentrations in the blood, the free base and maleate salt should produce the same results, and one would expect the same activity between the two forms. Nichols Tr. 2883:10-16; Tupper Tr. 505:11-17. Indeed, prior to trial, Dr. Reith made no distinction between compound 6 and its free base (compound ’222) and actually treated them interchangeably. Reith Tr. 909:8-20.

161. Accordingly, one of ordinary skill in the art would not have understood from a close reading of

Chakrabarti 1980a

that compound 6 was a preferred compound, and hence, that a hydrogen in the 7-posi-tion would have been preferred.

(b) Compound 36

162.

Chakrabarti 1980a

reported that compounds 34, 35, and 36 “retained good activity.” ' TX 3465 at 879, col. 2 (“However, compounds (34-36) ... retain good activity.”). The statement regarding the activity of compounds 34-36 was made in the context of comparing the compounds, in which each had a hydroxyethyl or hy-droxypropyl on the piperazine ring (and which had good activity), to compounds 31-33 having other substituents on that ring (and which were less active or inactive). Nichols Tr. 2830:18-2831:14; TX 3465 at 879.

163. Compound 36 received a CAR score of two at a dose of 10 mg/kg, and a

*849

CAT score of two at a dose of 12.5 mg/kg. As stated previously, a CAR score of two was not preferred. Pullar Tr. 227:1-22, 255:12-256:18.

164. Dr. Tupper testified that because of the manner in which the test was scored, a reported CAR score of two encompassed a broad range of activity ranging from a 31% to 50% block of the conditioned avoidance response. Pullar Tr. 229:12-230:14; TX 3465 at 882 n.d. Only the authors knew where in that score range each compound fell, and there is no way for the reader to second guess the judgment of the authors as to which compounds were “good” and which were not. Pullar Tr. 294:15-24; Nichols Tr. 2778:12-22; LaVoie Tr. 1561:17-25.

165. Therefore, the fact that- the authors stated that compound 36 “retained good activity” with a CAR score of two can at most be interpreted as a reference to a compound at the high end of the “two” range approaching a 50% block. It cannot be taken as a statement that all compounds with a reported CAR score of two (with activities as low as 31%) were good, active compounds when the remaining text of the article indicated that they were not.

166. One of ordinary skill in the art would not have understood from a reading of

Chakrabarbi 1980a

that compound 36 was a preferred compound. The fact that the authors’ stated that compound 36 “retained good activity,” read in its proper context, is not evidence to the contrary.

(c) Compound 34

167. Compound 34 is singled out by the authors as a preferred compound. TX 3465. This compound, like olanzapine, has a hydrogen substituent at the Rx or 7-position, but, unlike olanzapine, also has a hydroxyethyl piperazine group on the pi-perazine ring. Such hydroxyethyl pipera-zine groups were known to enhance the activity of antipsychotic drugs. LaVoie Tr. 1564:24-1565:4, 1567:11-22; Nichols Tr. 2779:25-2786:22 (“[T]here was a recognition in the art that a hydroxyethyl had special properties ... [I]n general, the compounds were more potent. They usually had better brain penetration ... ”).

168. The identification of five preferred compounds by

Chakrabarbi 1980a,

each of which had either a fluorine or a hyroxye-thyl piperazine substituent, could not rationally be read by a person actually skilled in this field to be an expression of preference for compounds like olanzapine that have neither a fluorine atom nor a hydroxyethyl piperazine group. Nichols Tr. 2786:12-2788:8; Reith Tr. 901:15-902:8.

169. Nothing in

Chakrabarbi 1980a

provides motivation for a person with ordinary skill in the art to begin with compound ’222 over one of the compounds the authors endorse as.preferred. Nichols Tr. 2272:22-23, 2273:1-12.

b. Changing the 2-ethyl in Compound ’222 to a 2-methyl

170. One skilled in the art would not have been motivated to modify compound ’222 by changing the ethyl group to a methyl group at the 2-ppsition of the thiophene ring because there is no suggestion in the ’574 patent or any other prior art that such a modification would have increased activity and decreased unwanted side effects. Reith Tr. 915:20-916:1; La-Voie Tr. 1562:1-6; Nichols Tr. 2772:18-2774:7. Even if a person of ordinary skill in the art were motivated to begin with compound ’222, he would modify it by adding a halogen atom to give it the neuroleptic substituent.believed at the time to be required for antipsychotic activity. Nichols Tr. 2773:13-19.

171. Although

Chakrabarbi 1980a

reports that “[a] short alkyl substitution ( [methyl, ethyl, isopropyl]) at position 2 of

*850

the thiophene ring seems to increase the activity,” TX 3465 at 879, col. 2, the article does not state that any particular short alkyl group will provide better results than any other short alkyl group. Reith Tr. 915:20-916:1; LaVoie Tr. 1574:2-4; Nichols Tr. 2273:20-2274:7.

172. The only compounds for which data is given allowing comparison of hydrogen at the 2-position with all of the “short alkyl” substitutions at the 2-posi-tion (as well as the longer alkyls, e.g., “bulky t-Bu group”) as described in the text are compounds 8 through 21 — all of which have a fluorine at the 7-position of the benzene ring. Tupper Tr. 435:20-25; Reith Tr. 915:5-19; LaVoie Tr. 1574:2-9; TX 3465 at 880-81. Compounds 6 and 7, neither of which is identified as preferred, provide an isolated example of a comparison between a short alkyl substitution and a t-butyl substitution in unfluorinated compounds. Tupper Tr. 487:1-7.

173. Regardless of this one comparison, the paragraph in

Chakrabarti 1980a

in which this preferred substituent is revealed states a preference for short alkyl groups in the context of a fluorinated series of compounds; thus the article does not generally provide motivation or guidance relating to the effect of such substitution on fundamentally different unfluorinated molecules. Reith Tr. 915:5-19; Nichols Tr. 2772:18-2774:7.

c. Replacing the Fluorine Atom in Flumezapine with a Hydrogen Atom

174. Defendants contend that one of ordinary skill in the art would have been motivated to replace the fluorine atom in flumezapine with a hydrogen atom in the 7-position of the benzene ring to arrive at olanzapine.

175. In light of the general state of the art, including the teachings of the ’574 patent and

Chakrabarti 1980a, Chakrabar-ti 1982,

and

Chakrabarti 1989,

one of ordinary skill in the art would have expected that replacing the fluorine atom with a hydrogen atom would produce a compound without sufficient antipsychotic activity. Nichols Tr. 2776:5-11.

176. While

Chakrabarti 1980a

suggests that a chlorine atom in place of the fluorine atom would also enhance the compound’s activity, it does not specifically suggest that the same result could be obtained with a hydrogen atom. Nichols Tr. 2779:17-24; TX 3465 at 879, col. 2. Nor does anything in

Sullivan and Franklin

suggest the desirability of using a hydrogen atom at this position. Nichols 2776:5-11; TX 3161; Findings of Fact § IV.B.l.d. If one were looking to replace the fluorine, one would replace the fluorine with other electronegative groups, not hydrogen. TX 1315 at 3172; LaVoie Tr. 1572:12-1573:18. Indeed, the art as a whole teaches directly away from using hydrogen because it is not an electron-withdrawing substituent. Nichols Tr. 2773:3-12.

177. Moreover, the 1990 publication, Davis et al.,

Chloro-Substituted, Sterically Hindered 5,11 Dicarbo Analogues of Clo-zapine as Potential Chiral Antipsychotic Agents,

J. Med. Chem, 809 (1990), proposed that the possible toxicity of candidates that were being investigated as safe alternatives to clozapine (including flumezapine) might be due to the presence of so-called “hetero atoms” — nitrogens and sulfurs — in the three-ring system of those compounds. Nichols Tr. 2757:21-2758:10; TX 1356 at 809. Those hetero atoms have nothing to do with the fluorine atom, and if those hetero atoms were removed, one would not obtain olanzapine. Nichols Tr. 2757:21-2758:16; TX 1356.

5. Reasonable Expectation of Success

178. By April 1990, a person with ordinary skill in the art knew facts that would

*851

have foreclosed a reasonable expectation of success with other clozapine-like compounds. These include: (1) the first known atypical antipsychotic drug, cloza-pine, was known to interact with a large number of receptors in the brain, including dopamine receptors, serotonin receptors, and cholinergic receptors, Paul Tr. 121:20-122:8, LaVoie Tr. 1541:24-1542:8, Nichols Tr. 2763:6-2765:10; (2) it was not known through what combination of these or other receptors clozapine exerted its atypical antipsychotic action, Nichols Tr. 2763:6-16; (3) it was not known why clozapine caused agranulocytosis in humans, Paul Tr. 121:14-17, Nichols Tr. 2769:13-17; (4) many compounds identified in patents and publications during the 1970s and 1980s as potential antipsychotic drugs had failed in clinical trials due to lack of efficacy, lack of atypicality, and/or an array of adverse side effects, Nichols Tr. 2751:1-2757:14; TX 1356; (5) small structural changes in cloza-pine-like molecules yielded great and unpredictable changes in properties, Pentel Tr. 1926:25-1927:9; Nichols Tr. 2796:6-2804:14; and (6) many researchers made a large number of compounds over a long period of time in an effort to find a safe, atypical antipsychotic drug and failed. Nichols Tr. 2743:14-2744:17, 2751:1-4; TX 1397 at ZY 19 66.

179.In 1990, a skilled artisan would have believed that a halogen atom on clo-zapine was important for its antipsychotic activity, that the particularly active members of the thienobenzodiazepine family had a halogen, and that the only member of that family of compounds to be advanced to clinical trials, flumezapine, had a halogen atom. Nichols Tr. 2748:11-2750:9, 2772:18-2773:2; Reith Tr. 916:2-16; La-Voie Tr. 1566:5-12; TX 3129; TX 3131; TX 3132; TX 3465. A person of ordinary skill in the art would not likely assume that the halogen substituent widely used on this class of compounds could be eliminated without adverse effect. Nichols Tr. 2774:8-19.

180. In 1990, a skilled artisan would not have reasonably expected, based on the data from the animal tests, that any of the compounds identified in

Chakrabarti 1980a

would be effective antipsychotics. The animal behavioral tests reported in

Chakrabarti 1980a

were only indicators of

'potential

antipsychotic activity. Reith Tr. 929:22-930:17; Nichols Tr. 2768:16-2769:11.

181. Moreover, these animal tests were not indicative of a compound being a safe, atypical antipsychotic, which was the goal of researchers after the toxicity of cloza-pine was recognized in patients. Pullar Tr. 173:21-174:12; TX 1031 at ZYP 506 1725; Reith Tr. 930:5-932:12; Nichols Tr. 2768:16-2769:24; Tupper Tr. 457:20-458:2. Indeed, the two most preferred compounds identified as showing good activity in the

Chakrabarti

prior art, flumezapine and ethyl flumezapine, turned out to be toxic. Pullar Tr. 180:2-181:8, 188:13-189:14. Thus, there would have been no reasonable basis from

Chakrabarti 1980a

to expect that new compounds such as olanzapine would have been safe, atypical antipsychot-ics.

182. In light of these facts, at the time olanzapine was developed, there could have been no reasonable expectation of success that an unhalogenated compound structurally similar to compound ’222 would succeed as a safe and effective atypical antipsychotic.

6. Composition and Method Claims

183. Claims 2, 3, 7, 8, and 15 are the composition and method claims of the ’382 patent. TX 1000.

a. Claim 1 claims the compound olanza-pine. TX 1000, col. 12.

*852

b. Claims 2, 3, and 15 claim “pharmaceutical compositions” of the compound claimed in claim 1. TX 1000, col. 12.

c. Claim 7 claims a method for treating an animal suffering from or susceptible to schizophrenia using the compound claimed in claim 1. TX 1000, col. 12.

d. Claim 8 claims “[a] method of claim 7 wherein the effective amount is from 0.1 to 20 mg per day” of the compound claimed in claim 1. TX 1000, col. 12.

184. The ’574 patent, the

ChakrabaHi

publications, and the

Sullivan and Franklin

article do not teach or suggest either the use of olanzapine in amounts less than 20 midday to treat schizophrenia or the dosage forms for such treatment. Dr. Reith and Dr. LaVoie did not offer any trial testimony that the references teach or suggest what an “effective amount” of olanzapine would be to treat a human suffering from or susceptible to schizophrenia. Dr. Reith agreed that the ’574 patent refers to doses in the range of 0.1 to 20 mg/kg/day, equivalent to between 7 and 1400 mg/day for a 70 kilogram (150 pound) person, and that the ’574 patent provides no guidance as to where in that range one should treat an actual schizophrenic patient with olanzapine. Reith Tr. 905:8-906:3. Similarly, Dr. LaVoie conceded on cross-examination that the ’574 patent referred to 0.1 to 20 mg/kg/day while the ’382 patent referred to 0.1 to 20 mg/ day, a 70-fold difference for a 70 kilogram (150 pound) person. LaVoie Tr. 1574:23-1575:9.

7. Secondary Considerations

a. Long-Felt Need

185. Beginning at least as early as 1975, there was a long-felt need for a safe, atypical antipsychotic drug that remained unsatisfied at the time Lilly filed the olan-zapine patent application in 1990. Nichols Tr. 2751:1-4, 2808:5; Schultz Tr. 2971:8-2972:17; TX 1397 at ZY 19 66 (“The medical need for better antipsychotic drugs in terms of increased efficacy and fewer unwanted effects is great”).

186. Numerous investigators, including scientists at Lilly, tried but failed to develop a safe, atypical antipsychotic drug between 1975 and 1990. Nichols Tr. 2751:1-12; Schulz Tr. 2971:8-2972:17; TX 1356; TX 1397 at ZY 19 46-48, 54-58, 66.

187. Risperidone (marketed and sold by Jansen Pharmaceuticals as Risperdal) is an atypical antipsychotic. It was first prescribed to schizophrenic patients in February 1994. Olanzapine (marketed as Zyprexa) was first prescribed in October 1996. Schulz Tr. 2987:15-17, 2988:15-17; Kinon Tr. 2526:24-2527:2; TX 1590; LD 113.

188. Risperidone and olanzapine are prescribed by doctors more than any other atypical antipsychotic on the market.

See

TX 1590. Risperidone, however, is prescribed more often than olanzapine. TX 1590 at ZYP 528 16; LD 113; Paul Tr. 150:13-151:2;

see also

Finding of Fact #193.

189. There is no evidence in the record to establish when the risperidone patent application was filed.

190. Because risperidone was not prescribed or otherwise available to schizophrenic patients at the time the ’382 patent was filed, olanzapine met the long-felt but unsolved need for a safe, atypical antipsy-chotic.

b. Failure of Others

191. As reflected in the court’s previous findings, there was a failure by others to develop a safe, atypical antipsychotic drug prior to the filing of the olanzapine

*853

patent application in the U.K.

See

Findings of Fact § II.D.

c. Commercial Success

192. Since late 1996, when olanzapine was approved for use, it has captured significant market share in terms of the number of prescriptions written for antipsy-chotic medications. TX 1590.

193. In 2001, olanzapine accounted for more than 25% of all antipsychotic prescriptions written, and risperidone accounted for more than 29% of all antipsy-chotic prescriptions written. TX 1590.

194. In 2001, olanzapines’ U.S. sales were $2.18 billion, representing one third of Lilly’s revenues for that year. TX 1597; Paul Tr. 139:6-8.

195. Lilly also spent a substantial portion of its revenue on marketing and administrative costs. Between 1996-2001, Lilly spent more on marketing and administrative costs than on research and development. Paul Tr. 148:5-149:3.

196. The evidence of the commercial success of olanzapine does not weigh in favor of either Lilly or the Defendants.

d. Industry Acclaim

197. Olanzapine has received substantial industry acclaim in the form of the Prix Galien Award in 1997, and Queens Award for Enterprise in 2000, and the Pharmaceutical Manufacturer’s Association Discoverer’s Award in 2000. In addition, testimonials from treating doctors and patients provide objective evidence of olanzapine’s nonobviousness. Trial Tr. 2730:12-2731:15 (stating the February 10, 2004 Stipulation regarding awards received for the discovery of olanzapine); Paul Tr. 142:4-143:8; Schulz Tr. 2973:10-2974:2.

198. Although the inventor of risperi-done received the Prix Galien Award in 1996, Trial Tr. 2730:24-25, that fact does not vitiate the industry acclaim held by olanzapine.

e.Unexpected Differences Between Compound ’222 and Olanzapine— The Dog Studies

(1) Basic Principles of Toxicology

199. Drug development relies heavily on animal testing to try to find drugs that will be safe for human use. Testing first directly on human subjects was, and still is, unethical and unheard of in modern drug development. Instead, testing in animals always precedes human clinical studies. The genetic make-up, organ systems, and biochemistry in other mammals are sufficiently similar to those in humans that one can study adverse effects of drugs in laboratory animals. Dr. Shayne Gad (“Dr. Gad”), Zenith’s expert in toxicology, testified, “[i]t is á fundamental hypothesis of toxicology that adverse effects caused by chemical entities in animáis are generally the same as those induced by those entities in humans ...” Gad Tr. 1753:9-13. Thus, the field of toxicology relates generally to the testing of drugs in animals in an effort to identify toxic risks for humans. Emmerson Tr. 512:17-20; Gad Tr. 1754:1-12.

200. Toxicity testing is done in order to identify the potential toxic effects of drug candidates under development. Gad Tr. 1754:1-6, 1756:1-12. Such testing is done at doses chosen to elicit any toxic effects, that is, at or near the “maximum tolerated dose” or “MTD:” Maximum tolerated dose means the highest dose an animal can receive consistent with living and general good health. Emmerson' Tr. 539:4-8. The MTD and fractions thereof are appropriate doses for pharmaceutical toxicology testing so that the full range of potential toxicities can be observed. Emmerson Tr. 539:10-14; Kanter Tr. 2102:9-2103:4. Indeed, if a compound is not tested at or near the

*854

MTD, an “important adverse finding that is attributed to treatment with the compound” can be missed. Emmerson Tr. 539:22-25. This is because humans are generally more sensitive to the toxic effects of drugs than a homogeneous population of laboratory animals. Emmerson Tr. 526:4-13, 539:15-25.

(2) Lilly’s and Defendants’ Dog Toxicology Studies

(a) Lilly’s D07290 Dog Study

201. The D07290 Dog Study was a toxicology study conducted by Lilly at the request of Lilly’s patent department. Em-merson Tr. 568:3-7.

202. The D07290 Dog Study compared five groups of beagle dogs: a control group that received a placebo, a group that received 4/mg/kg/day of olanzapine, a group that received 4 mg/kg/day of compound ’222, a group that received 8 mg/ kg/day of olanzapine, and a group that received 8 mg/kg/day of compound ’222. Each group consisted of eight dogs, four males and four females. TX 1164 at ZY 622 340.

203. Drug administration to dogs commenced on September 25, 1990, and ended on April 2, 1991. TX 1164 at ZY 622 340.

204. At the conclusion of the D07290 Study, Lilly claimed that the four females in the high-dose compound ’222 group showed a statistically significant increase in group average (mean) cholesterol over time as compared to female control animals and the high-dose female dogs given olanzapine.

See'TK

1001.1 at FH 150-51, 154.

(b) Zenith’s Dog Study — The MPI Study

205. Zenith commissioned a dog toxicology study, which was conducted at MPI Research, Inc. (“MPI”), for purposes of this litigation (the “MPI Study” or “Zenith Study”). The Zenith Study, like the Lilly Study, lasted six months. Drug administration to dogs lasted from May 3, 2002, to November 4, 2002. TX 1188 at ZG 001604.

206. The Zenith Study was an expanded comparison of olanzapine and compound ’222 in female beagle dogs at a dose of 8 mg/kg/day. TX 1188 at ZG 001608. Included in the Zenith Study were measurements of progesterone in order to evaluate the estrous state of the female dogs. TX 1188 at ZG 001613. Zenith’s Study design thus allowed consideration of the estrous cycle in individual animals and permitted study of the possible effect of the drug on that cycle as well as the relation of that cycle to the levels of cholesterol that were observed. Nachreiner Tr. 1099:6-25. The Zenith Study was conducted according to GLP guidelines. TX 1188 at ZG 001603.

207. In the Zenith Study, there were fifteen female dogs assigned to each of three groups. Group 1 was a nontreated control group. Group 2 was a group in which each dog was given 8 mg/kg/day of ’222. Group 3 was a group in which each dog was given 8 mg/kg/day of olanza-pine. TX 1188 at ZG 001608. The dogs were assigned to the groups based on body weight and cholesterol levels, but not based on pretest estrous state. TX 1188 at ZG 001606. Blood samples for cholesterol analysis were taken before the test began, twice within the first week after dosing started, and then at monthly intervals. TX 1188 at ZG 1611-12.

208. Before administering the test compounds, Zenith divided the dogs into three test groups that all had the same average cholesterol level, excluding in the process any dog with a cholesterol level above 175 mg/dl. Nachreiner Tr. 1111:2-

6;

Goldenthal Dep. 131:8-15. The dogs in the Lilly D07290 Study were not preselected to have cholesterol levels under 175 mg/dl. Nachreiner Tr. 1111:7-9. The dif

*855

ference between the average cholesterol values of the untreated dogs in the Zenith Study and the untreated dogs in Lilly’s D07290 Study was approximately 40 mg/dl. Nachreiner Tr. 1112:8-11.

209. The numerical magnitude of the increase in cholesterol caused by compound ’222 in both the D07290 and Zenith Studies was about 40 mg/dl. Thisted Tr. 3108:5-3109:6. When the effect of progesterone (estrus) is filtered out, the magnitude of the estimated difference is about the same, i.e., about 40 mg/dl. Thisted Tr. 3093:25-3095:3; Gibbons Tr. 2221:10-22, 2269:8-11. The numerical difference in mean cholesterol levels, 40 mg/dl, and the fact that the mean cholesterol was shifted upward by one standard deviation, are both good measures of effect magnitude. Thisted Tr. 3122:3-9.

210. The Zenith Study also purported to measure the fractions of total cholesterol carried as HDL and LDL. Beginning with the week two bleeding, and at every subsequent time point, the mean LDL cholesterol values for the compound ’222-treated dogs were statistically significantly higher than those for the olanzapine-treat-ed dogs and the control dogs. Nachreiner Tr. 1118:12-1119:3; Goldenthal Dep. 265:1-11. Over the course of six months, the compound ’222-treated dogs experienced an 18 point increase in HDL and a 22.6 point increase in LDL. Scanu Tr. 1277:13-1279:6,1309:12-14. In these dogs, the percentage of LDL increased six times more than the percentage of HDL. Scanu Tr. 1309:5-11.

211. In short, the results of Zenith’s MPI Study confirmed the findings of the D07290 Study that cholesterol in the female high-dose compound ’222 group was significantly greater than in the olanzapine and control groups, whereas the olanza-pine group did not differ significantly from the control group.

(c) DRL’s Dog Study — The Calvert Study

212. Unlike the Lilly and Zenith' toxicology studies, the study commissioned by DRL and conducted by Calvert Preclinical Services, Inc. (the “Calvert Study”) lasted only sixty days — from April 2, 2002, to July 2, 2002. Gayheart-Walsten Dep. 83:16-18; TX 1180 at DRL-IN 055560, 055563. The only clinical chemistry variable studied was cholesterol. TX 1180 at DRL-IN 055560. There were ten female beagle dogs , in each of four groups: (1) a control group, (2) a group given 8 mg/ kg/day of compound ’222, (3) a group given 8 mg/kg/day of olanzapine, and (4) a group of untreated dogs that were offered double rations of food. The researchers gave the dogs in Group 4 double the normal ration of food to see if doing so would have an effect on cholesterol concentrations. TX 1180 at DRL-IN 055560, 055566-67.

213. WTiile the Calvert Study is fundamentally too short to compare to the longer Lilly and Zenith Studies, Rebar Tr:. 2318:22-2319:2, the observed effects ' on cholesterol are consistent with those in the longer studies. Thisted Tr. 3124:20-3126:9. For example, there was a rapid increase in cholesterol concentrations in the compound ’222-treated group. Thist-ed Tr. 3125:24-25. The mean cholesterol values in the compound ’222-treated dogs were' greater than in the olanzapine-treat-ed dogs at all time points. Kanter Tr. 2162:22-2163:5; Gayheart-Walsten Dep. 102:5-17. In addition, the cholesterol levels of the control dogs rose over the course of the Calvert Study. Thisted Tr. 3126:1-3.

214. The cholesterol concentrations did not increase in the group of dogs offered double rations of food. Gayhearh-Walsten Dep. 103:13-20, 104:8-105:12; Kanter Tr. 2160:16-18.

*856

(3) Criticisms of Lilly’s Dog Study Evidence

215. Defendants attack the D07290 Study on numerous grounds and assert that the data and the conclusions drawn therefrom by Lilly scientists are seriously confounded.

(a) The Dog as a Model

216. The dog is an appropriate species in which to test for potential toxic effects in humans, including effects on total cholesterol. Emmerson Tr. 525:4-18;

see, e.g.,

TX 1272; TX 1278; TX 3072 at DRL IN 055565; TX 3087.

217. By 1990, the beagle dog had emerged as the large mammal of choice for toxicology testing of new drug candidates. Emmerson Tr. 572:3-19; Means Tr. 1998:9-18. The dog, while not perfectly predictive, is reasonably predictive of toxic effects in humans. Gad Tr. 1754:1-6; Ranter Tr. 2105:8-20, 2120:16-22; Means Tr. 1998:24-1999:7. A wealth of expert testimony and literature supports the finding that the dog is an acceptable toxicity model for humans. The writings of Zenith’s expert toxicologist, Dr. Gad, teach “that the dog is currently the first choice nonrodent model for toxicity studies ...” Gad Tr. 1758:1-4.

218. The dog model was widely used by Lilly in a variety of studies early in the development work leading to the discovery of olanzapine. These studies, conducted and reviewed over a period of years, provided the basis for making decisions to terminate or advance testing of drug candidates such as olanzapine.

See

Emmer-son Tr. 542:6-21, 543:10-13, 545:17-546:5, 546:22-547:1.

(b) Total Cholesterol as a Tested Parameter

219. Over the years, a standard battery of observations and measurements seeking to detect the toxic effects of drugs in dogs has been developed. As technology and measurement methods improved, this standard battery grew. By 1990, the standard battery included measurement of total cholesterol. Gad Tr. 1758:5-13; TX 3107 at 121; Emmerson Tr. 536:16-23. The design of the D07290 Dog Study, therefore, included a measurement of total cholesterol as part of the then standard battery of dog toxicology tests. Emmerson Tr. 535:16-536:25.

220. In this regard, the laboratory retained by DRL to conduct DRL’s Calvert Study for this litigation reported that “[t]he beagle dog is an acceptable species to study the effects of compounds and diet on total cholesterol, triglycerides, HDL, LDL, and VLDL.” TX 3072 at 11. Similar conclusions have been drawn in the literature. A 1997 article in the journal

Atherosclerosis

states in reference to the beagle dog, “[t]his animal model has been successfully used to demonstrate the effect of statins [cholesterol lowering drugs] on cholesterol ...” and that “this animal model has good predictive power for hypoeholest-erolemic effects in man.” TX 3087 at 203; Scanu Tr. 1286:20-1288:10; Davidson Tr. 2919:2-14.

221. Both dogs and humans can experience elevations in total cholesterol concentrations as the result of similar metabolic effects and abnormalities. Scanu Tr. 1260:11-1263:24; Gad Tr. 1810:16-22. Examples of hypercholesterolemias (elevation of serum cholesterol concentrations) in both dogs and humans with common etiologies include the ingestion of high-fat diets, cholestasis, hypothyroidism, diabetes mellitus, pancreatitis, hyperadrenocorti-cism, and nephrotic syndrome. Bauer Tr. 2665:23-2668:15; TX 1347.

222. There is a known and reported nexus between dog studies and humans with regard to total cholesterol. Dr. Michael Davidson (“Dr. Davidson”), Lilly’s

*857

expert in cardiology, cholesterol, and cardiovascular disease, Davidson Tr. 2891:13— 14, 2897:25-2898:5, testified that he has been involved in human clinical trials for statin drugs, including Mevacor, Zocor, Pravacol, Lipitor, and Crestor. Davidson Tr. 2898:17-2894:5. Dr. Davidson testified that the experimental statins were tested in dogs before they were tested in people. Davidson Tr. 2919:2-21. Indeed, the dogs proved to be a reasonable predictor of cholesterol effects in humans in that sta-tins decrease cholesterol in both dogs and humans. Davidson Tr. 2918:12-15, 2919:2-21. Moreover, there are a number of articles regarding statin research that show that the dog is used by large, established pharmaceutical companies as an appropriate model to indicate whether drugs may lower cholesterol in people.

See, e.g.,

Scanu Tr. 1286:20-1293:10; TX 1272, 1273, 1275, 1278, 3087. Dr. Davidson further testified that rapamycin, a.k.a. Rapamune, a drug that prevents transplant rejection in organ transplant patients, raised cholesterol in both dogs and humans. Davidson 2926:1-5, 2927:12-14 (humans), 2928:16 (dogs);

see also

Scanu 1253:18-25 (humans); TX 1353 (humans).

223.Dr. John Bauer (“Dr. Bauer”), Lilly’s expert in lipids and lipoproteins in dogs as well as comparative studies in humans and dogs, Bauer Tr. 2655:10-12, testified that “observations of total blood cholesterol or total serum cholesterol [in dogs] using the terms equally are a useful predictive index for the response of total blood cholesterol in humans.” Bauer Tr. 2655:15-2659:4. This opinion is implicit in many of his writings and is explicitly stated in an article he wrote in 1996. Bauer Tr. 2655:25-2657:16; TX 1347. Dr. Bauer summarized three reasons for his opinion that dogs are good models to predict cholesterol effects in humans: (1) the many similarities in cholesterol metabolism in humans and dogs; (2) the similar effects in cholesterol in humans and dogs caused by diseases; and (3) the considerable literature on the effects of experimental compounds on cholesterol in dogs and humans, showing that cholesterol reductions or elevations seen in dogs were also seen in humans. Bauer Tr. 2658:2-2659:4.

224. There are recognized differences in the way cholesterol is metabolized in dogs and humans. Humans carry most of their cholesterol in LDL, the so-called “bad cholesterol,” while dogs carry most of their cholesterol in HDL, the so-called “good cholesterol.” Scanu Tr. 1210:6-17; Bauer Tr. 2708:18-22. Consequently, when total cholesterol is elevated in dogs, the resulting increase in HDL cholesterol does not usually form atherosclerotic plaque in dogs. In contrast, when total cholesterol is elevated in humans, it is carried primarily as LDL, thereby increasing atherogenic risk in humans. Bauer Tr. 2667:5-12, 2721:5-12.

225. Because dogs are resistant to elevated cholesterol while humans are not, Dr. Bauer opined that if an experimental compound elevated cholesterol levels in dogs, then he would expect either an equal or greater increase in humans. Bauer Tr. 2675:13-2676:17, 2678:4-7, 2678:18-2679:12. His opinion, although not supported by a published study or subjected to peer review, was based on his extensive experience studying comparative hyper-cholesterolemia in both species. Bauer Tr. 2703:15-2704:25. Thus, Dr. Bauer’s expression of this untested theory does not diminish his credibility as a witness in his area of expertise — lipids and lipoproteins in dogs and comparative studies in humans and dogs.

226. One of the articles which Dr. Bauer cited in his expert report is an article entitled “SR-12813 lowers plasma cholesterol in beagle dogs by decreasing cholesterol biosynthesis.” TX 3087. That article reports that the mechanism by

*858

which a statin drug works and the metabolic pathway through which it proceeds is known in both dogs and humans. TX 3087; Bauer 2694:12-2695:23. The mechanism by which a dog’s cholesterol is raised is unknown. Bauer Tr. 2698:1-5. Despite this fact, this article remains consistent with the fact that there are similarities between the cholesterol metabolism in dogs and humans which render the dog a good predictive model for humans.

227. Dr. Bauer’s opinion is supported by a chapter he co-wrote in

The Clinical Chemistry of Laboratory Animals,

TX 3779, which concluded that dogs remain a good model for predicting total cholesterol effects in humans. Bauer Tr. 2652:3-2654:3, 2656:25-2657:16.

228. Dr. Angelo Scanu (“Dr. Scanu”), Zenith’s expert in lipoproteins, testified that in his experience in lipoprotein metabolism, his review of the D07290 Study, and his review of published literature in the area of lipoprotein metabolism, the dog is not a good animal model for predicting cholesterol results in human patients. The principle reason cited for his opinion is the fact that cholesterol is metabolized in a strikingly different manner in dogs as opposed to humans. Scanu Tr. 1200:14-1201:2.

229. Contrary to Dr. Scanu’s opinion, the evidence demonstrates that this metabolic difference does not prevent the dog from being a good model for predicting total cholesterol results in humans. Bauer Tr. 2655:15-2659:4. Although differences exist between humans and dogs as to the manner in which total cholesterol elevations are metabolically transferred and disposed of, these differences are “downstream” from the causes of cholesterol elevations and, therefore, secondary to them. Bauer Tr. 2659:5-2665:18. For example, when dogs are fed high-fat diets, the hy-percholesterolemic response in dogs is to partition excess cholesterol into HDL fractions, whereas humans partition the excess cholesterol into LDL fractions. Scanu Tr. 1273:11-1274:16. In either event, however, both dogs and humans experience an elevation in total cholesterol. Bauer Tr. 2676:18-2679:14;

see also

Finding of Fact # 221.

(c) Randomization

230. . In a well-designed experiment, animals are randomly assigned to treatment groups to remove systematic effects (aside from the treatment being studied), known and unknown, that could affect the outcomes being measured. A purpose of randomization is to ensure that subjective judgments play no role in which animals are assigned to which treatment groups. This eliminates a source of bias that otherwise could systematically favor one treatment group over others for reasons unrelated to the treatment itself. Thisted Tr. 3082:15-3083:1.

231. Lilly’s design of a dog study using forty dogs randomly assigned by body weight is common in standard toxicology studies, and consistent with Lilly’s standard operating procedure. Emmerson Tr. 571:22-572:2; Gad Tr. 1765:15-21; Gayh-earh-Walsten Dep. 127:7-128:9. There are a variety of physiological • variations that might exist among dogs. The general purpose of random assignment based on body weight of dogs amongst the treatment groups is to balance out whatever variations may exist. ' Gad Tr.' 1765:18-21, 1766:6-9.

232. At the start of the D07290 Study, there was one dog in the compound ’222 group that weighed more than the others. The dog was not obese, and her weight was within the protocol guidelines. There is no evidence that the size of a dog affects its cholesterol levels. Rebar Tr. 2383:3 — 21; TX 1164 at ZY 622 607 (body weight 5-12 kg), 688-97.

*859

233. There was no reason that the dogs in the D07290 Study should have been randomized based on their estrous stage because (1) it was not common practice to examine female dogs for their estrous stage prior to standard toxicology tests,

see, e.g.,

Goldenthal Dep. 164:7-9 (stating that MPI has never randomized by estrous state prior to a study); (2) it was a common practice to randomize dogs based on body weight prior to standard toxicology tests; and (3) Lilly did not know at the outset that it would find a result that may be affected by the estrous stage. Gad Tr. 1764:8-15, 5:15-1766:9; Emmerson Tr. 571:22-572:2, 580:4-581:22. Even Zenith did not randomize by estrous state in conducting its dog study, even though it knew that cholesterol results would be very important. Goldenthal Dep. 162:15-166:13, 168:6-8.

(d) The Length of the Study

234. The length of the D07290 Study was “long enough to recognize a biological [sic] significant effect,” Rebar Tr. 2316:21-2317:1, i.e., to study the chronic effects of drug candidates and determine their suitability for repeated-dose studies in humans. Gad Tr. 1757:15-25; Rebar Tr. 2333:22-2334:20. Such preclinical six-month studies had been done at Lilly for both ethyl flumezapine and flumezapine. Emmerson Tr. 543:10-13, 544:23-545:4.

(e) The Number of Dogs and Analysis by Sex

235. The forty dogs that Lilly used in conducting the D07290 Study were an appropriate number of dogs to use. Rebar Tr. 2316:13-16, 2331:3-12, 2331:24-2332:14.

236. It is standard practice in the scientific community to test toxicity in both male and female dogs and to analyze the results separately by sex. Gad Tr. 1767:3-8;

see also

Pentel Tr.l914:6-10, 1916:17-1918:11 (agreeing this is common and appropriate); Emmerson Tr. 535:6-11. Indeed, it is required by the FDA. Davidson Tr. 2938:20-2939:14.

237.Lilly used four dogs per sex per group in' dog studies submitted to and accepted by the FDA. Emmerson Tr. 535:2-5, 547:15-18, 569:8-10; TX 1005 at ZYP 177 734; TX 1086 at ZYP 661 721; TX 1087 at ZYP 661 1010'. Lilly submitted to the FDA a three-month olanzapine study done in 1983, two one-year olanza-pine studies (done in 1984 and 1993), and the D07290 Study, that all used four dogs per sex group. Emmerson Tr. 547:15-18, 553:10-11. Lilly also used the same number of dogs in its preclinical testing of flumezapine, which the FDA approved for initial safety and efficacy testing in humans. TX 1005; TX 3421. Four dogs per sex treatment group is a common number of animals to use. Gayheart-Walsten Dep. 35:3-6; Spainhour Dep. 85:8-13; Selim Dep. 19:20-20:22. This sample size was appropriate to perform statistical analysis and achieve statistically significant results. Thisted Tr. 3061:16-3062:7; Symanowski Tr. 708:2-10.

(f)The Dosage Used

■238. At the time of the design for the D07290 Study, there existed considerable .experience at Lilly’s laboratories with the effects of chronic administration of olanza-pine and related compounds in dogs.

See

Engelhardt Dep. 34:3-11 (stating that the dosage for the D07290 Study was selected based on a previous comparison of flume-zapine and ethyl flumezapine). As explained above, toxicology testing is traditionally done at or near the MTD, i.e., at higher than therapeutic doses, because the science of toxicology accepts effects in dogs at doses greater than the human therapeutic dose to be reliable indicators of human toxic risk. Gad Tr. 1768:18-1769:5. Comparable toxicology studies done with ethyl flumezapine, flumezapine,

*860

and olanzapine prior to actual human testing of those compounds were done at 4, 8, and 12 mg/kg for ethyl flumezapine; at 1, 2, 4, 8, and 12 mg/kg for flumezapine; and at 2, 5, and 10 mg/kg for olanzapine.

See

TX 1002; TX 1003; TX 1004; TX 1005; TX 1040; TX 1041;

see also

LD 22., Thus, the evidence establishes that the 4 and 8 mg/kg doses of compound ’222 and olanza-pine selected for use in the D07290 Dog Study were well within the objectively reasonable range of preclinical toxicology doses for compounds of this type. Emmerson Tr. 569:4-7; Gad'Tr. 1779:24-1780:4; Ranter Tr. 2165:16-2166:18.

239. A one-week pilot study of 8 mg/kg of compound ’222 confirmed that the dogs could tolerate 8 mg/kg of compound ’222 and that therefore, it was appropriate to proceed with 8 mg/kg of compound ’222 as the high dose in the D07290 Dog Study. TX 1164 at ZY 622 344; Emmerson Tr. 569:11-570:1.

(g) The Use of Equal Doses of Olan-zapine and Compound ’222 ■

240. It was appropriate for Lilly to use equal doses of olanzapine and compound ’222 in the Lilly D07290' Dog Study. First, the available

in vitro

and

in vivo

pharmacological data for compound ’222 and olanzapine suggested that they would be relatively equally active.

See

TX 1001.1 at FH 125-34, FH 143-45; Tye Dep. 169:17-170:2, 170:24-171:11, 171:13-20 (and errata). Second, one of the objectives of the D07290 Dog Study was to test the hypothesis that these compounds were so similar in structure that they would have essentially the same properties. Nichols Tr. 2860:13-17. The reasonable way to achieve this objective was to test the compounds at equal doses. Killworth Tr. 3022:5-3023:12.

241. Defendants assert

the

dose

for

compound ’222 was inappropriate as too great a multiple of the human therapeutic dose, and that therefore, the cholesterol results observed in the D07290 Study have no practical relevance. Gad Tr. 1712:2-1713:4; Pentel Tr. 1893:12-17. However, there is no evidence of a therapeutic dose or safety margin for compound ’222 because compound ’222 has never been tested in humans. Pentel Tr.l910:6-10. Moreover, Defendants’ experts admitted that' dosing the dogs at 4 and 8 mg/kg of compound ’222 was reasonable given the fact that olanzapine had been tested at 2, 5, and 10 mg/kg. Gad Tr. 1779:24-1780:4; Ranter Tr. 2165:16-2166:18; LaVoie Tr. 1574:19-1575:16.

(h) How the Study Was Conducted

i) Good Laboratory Practices Were Followed

242. The D07290 Dog Study was carefully conducted in accordance with its protocol and with the FDA’s GLP guidelines. 21 C.F.R. § 58 ; Emmerson Tr. 570:2-4. Lilly maintained complete records. Data reports and logs show careful observation of the dogs and comprehensive consideration of the health of the animals. The D07290 Dog Study was actually reported to the FDA in connection with Lilly’s request for approval to market olanzapine. Emmerson Tr. 570:15-24; TX 1164.

ii) Double Rations

243. After observing significant weight loss during D07290 in Dogs 240584 and 242547, two of the female high-dose compound ’222 dogs, Lilly researchers gave them double rations of food. Dog 240584 ate her double ration of food 75% of the time. Nachreiner Tr. 1057:2-6, 1057:24-1058:11; TX 3437 at ZY 622 685. Dog 242547 ate her double rations every day of the study but one. Nachreiner Tr. 1057:9-14, 1058:12-22; TX 3437 at ZY 622 687.

244. It is standard toxicology practice to offer increased food rations to dogs to

*861

help them maintain their body weight and complete the study. Emmerson Tr. 579:23-580:22; TX 1179 at ZYP 570 152;

see also

Goldenthal Dep. 111:17-112:17 (testifying that MPI generally gives the dogs in studies as much food as they want and then measures the amount they eat); TX 3195.

245. After the double feeding began for these dogs, their cholesterol increased and never came back down to the point it was at before they were double-fed. Emmer-son Tr. 613:23-614:4; Symanowski Tr. 734:16-20; Nachreiner Tr. 1060:5-1061:2, 1063:10-21,1065:4-9.

246. Both dogs regained weight but were continued on double rations throughout the remainder of the D07290 Study. Nachreiner Tr. 1059:16-24.

247. Dr. Bauer testified that feeding dogs increased amounts of Purina 5007 dry dog food, the food offered to the dogs in the D07290 Study in order to maintain their normal body weight, “would have no effect on their blood or total serum cholesterol” because

that particular type of dry extruded dog food is one that contains only modest amounts of its total calories from fat. And feeding — simply feeding larger or increased amounts, even double amounts, of that type of dog food that’s very low in its total calories from fat to a dog in order to maintain its normal body weight will not cause an elevation in total blood cholesterol ... if one wanted to do that by diet, it would take feeding a diet which had a much greater percentage of its total calories from fat in order to elevate the cholesterol in the dog. It’s not a question of feeding more of a low or moderate fat diet; it’s a question of feeding a diet which contains higher amounts of the total calories from fat.

Bauer Tr. 2682:19-2683:16.

248. The two dogs offered double rations had the lowest cholesterol levels in their group. Symanowski Tr.' 706:18-707:8.

249. A similar finding was noted in DRL’s Calvert Study. Thisted Tr. 3161:1-9 (testifying that the cholesterol levels in the group of dogs offered double rations of food did not increase);

see also

Findings of Fact § IV.B.7.e(2)(c).

250. Dr. Ronald Thisted (“Dr. Thist-ed”), Lilly’s expert statistician, performed a robustness check removing the two dogs offered double rations, and found that the results remained statistically significant, Thisted Tr. 3085:4-24, as did Dr. Robert Gibbons (“Dr. Gibbons”), Zenith’s expert statistician, using a different method. Gibbons Tr. 2218:24-2220:7; TX 3236 at ZG 13034, 13037; Thisted Tr. 3085:25-3086:8.

251. Although offering double rations to two of the dogs in the D07290 Study caused a difference in the treatment of some of the tested animals which is not ideal,

see

Thisted Tr. 3083:2-3, the evidence establishes that the offer of double rations alone did not confound the results of the D07290 Study.

(i) The Results of Lilly’s and Zenith’s Studies

i) Statistics Experts for All of the Parties Found a Statistically Significant Cholesterol Increase in the Compound ’222-Treated Dogs

252. The statisticians for each of the parties analyzed the data from the D07290 Study and found the cholesterol effect associated with compound ’222 was statistically significant over the course of a chronic treatment period. Dr. Symanowski’s original repeated measures analysis showed that for cholesterol in the female dogs at the 8 mg/kg/day dose level (a) there was a difference over time between

*862

olanzapine and compound ’222 that was highly statistically significant, with a p-value of 0.001, (b) there was also a difference over time between compound ’222 and the untreated (control) group that was statistically significant (p< 0.001), and (c) there was no statistically significant difference between the olanzapine and control groups. TX 1001.1 at FH 149-54; Thisted Tr. 3079:4-15; Symanowski Tr. 644:12-645:10, 653:17-654:9. The results of Dr. Symanowski’s statistical analysis of the D07290 Study have been independently confirmed by both Lilly’s statistics expert, Dr. Thisted, and by the Defendants’ experts using a variety of methods. Thisted Tr. 3079:16-3082:11, 3085:5-3089:12; LD 391. Dr. McDougall, DRL’s expert, replicated Dr. Symanowski’s repeated measures analysis and obtained the same results. Thisted Tr. 3080:15-16. Zenith’s expert, Dr. Gibbons, used the mixed effect regression method and also found that the high-dose compound ’222-treated female group showed a statistically significant increase in mean cholesterol over time compared to both the olanzapine and -control groups, but no statistically significant difference between olanzapine and control. Thisted Tr. 3080:17-22; Gibbons Tr. 2214:14-18, 2215:10-2218:22, 2236:4-14; TX 3236 at ZG 12909; LD 391.

253.Dr. Thisted and Dr. Gibbons each analyzed the D07290 data with different modern techniques that allowed them to include both protocol and nonprotocol measurements for all of the dogs. They still found a statistically significant effect of compound ’222 in the female high-dose group. Thisted Tr. 3080:23-3082:1; Gibbons Tr. 2216:19-2217:1, 2218:18-22. They also found statistical significance when they excluded data for the two dogs from the compound ’222 group that received extra rations. Thisted Tr. 3085:4-24; Gibbons Tr. 2218:24-2219:24.

254. The results of Zenith’s MPI Study confirmed the findings of Lilly’s D07290 Study which showed that cholesterol in the female high-dose compound ’222 group was significantly greater than in the olanzapine and control groups, whereas the olanza-pine group did not differ significantly from the control group. The statistical significance of the results of the MPI Study were found using several different statistical tests. Thisted Tr. 3089:13-3097:19; LD 17; LD 131; LD 381(b); LD 390. Dr. Gibbons confirmed these results as did Dr. Thisted using the GEE analysis. Thisted Tr. 3089:21-3090:12; Gibbons Tr. 2204:3-13. Dr. Thisted also looked at the last four months of the MPI Study, when cholesterol levels stabilized, and found a highly statistically significant difference between compound ’222 and both olanzapine and control during that period. Thisted Tr. 3090:13-3092:3; LD 17. MPI’s own data tables of the study results show a difference between compound ’222 and olanzapine throughout the study (using a statistical test called a t-test), and that compound ’222 was statistically significantly greater than control at months two, four, and five (using Dunnett’s test). TX 3076; Thisted Tr. 3092:11-3093:4. When the effect of estrus on cholesterol levels in the control group is taken into account, the difference between compound ’222 and control was statistically significant at the three- and six-month time points as well. Thisted Tr. 3095:4-3097:19; LD 381(b). Over the last three months of the MPI Study, the difference between the mean cholesterol values, in the compound ’222 group and the olanzapine group was about 40 mg/dl, a result also seen in the Lilly D07290 Study. Thisted Tr. 3108:5-3109:6; Gibbons Tr. 2269:8-11.

255. Furthermore, the availability of progesterone measurements in the MPI Study makes it possible to separately evaluate the effects of drug treatment and

*863

progesterone on cholesterol. Based on the method of evaluating the progesterone effect set forth by Dr. Raymond Nachreiner (“Dr. Nachreiner”), Zenith’s expert in veterinarian endocrinology and reproductive physiology, and based on his own study of the graphs of the individual dogs, Dr. Thisted performed an analysis of the MPI Study data and found that when progesterone effects on cholesterol are filtered out, there is still a highly significant increase in cholesterol in dogs given compound ’222. Thisted Tr. 3093:5-3094:12, 3161:16-22. In particular, Dr. Thisted found that the estimated magnitude of the cholesterol elevation, about 40 mg/dl, is about the same whether or not the effect of progesterone is taken into account. Thisted Tr. 3093:5-3094:12. The statistical model that he used, correcting cholesterol for the effects of elevated progesterone one month earlier, was a model proposed by Dr. Nachreiner. Thisted Tr. 3161:16-22.

256. Dr. Gibbons performed a similar analysis using a different method and also found that after the effect of the estrous cycle on cholesterol was filtered out mathematically, a statistically significant effect of compound ’222 remains, with the magnitude of 41 mg/dl. Gibbons Tr. 2220:11-2221:22, 2269:8-11; Thisted Tr. 3094:13-3095:3, 3161:23-3162:4. In addition, he found no statistically significant difference between the olanzapine and control groups. Gibbons Tr. 2221:23-2222:7.

257. Although the shorter DRL Study did not achieve a statistically significant difference between the compound ’222-treated dogs and the other groups at the p>0.05 level, the results are not inconsistent with the results of the six-month D07290 and Zenith Studies. Thisted Tr. 3124:20-3126:9; LD 117. All three studies exhibit the same pattern in the first two months, as is shown in LD 117, with a rapid rise in cholesterol in the compound ’222-treated dogs and a lesser increase in the olanzapine dogs when compared to the control group. Kanter Tr. 2138:5-16, 2161:20-2163:5. There is no way to tell whether the cholesterol levels in the compound ’222-treated dogs in the DRL Study would have stabilized after two months at a significantly higher level than those in the olanzapine and control groups, as was seen in Lilly’s D07290 Study and Zenith’s Study. Thisted Tr. 3124:20-3126:9.

ii) The Repeated Measures Analysis Was Appropriate

258. The technique used by Dr. Syma-nowski to analyze the results of the D07290 Study was the repeated measures analysis of variance. Symanowski Tr. 678:22-679:4, 692:13-19. Dr. Symanowski performed the analysis himself using the statistical analysis system (“SAS”) software to directly analyze data in the electronic database. Symanowski Tr. 674:15-675:10.

259. The repeated measures method was more appropriate than the Dunnett’s test method for comparing toxicity profiles of olanzapine and compound ’222, as was called for originally in the protocol. Sym-anowski Tr. 682:5-20, 2040:4-9; Thisted Tr. 3059:17-23, 3078:11-19; TX 3439 at ZYP 187 719. The repeated measures method has greater statistical power for detecting differences between treatment groups than does the Dunnett’s test. Symanowski Tr. 679:20-680:2. At the time Dr. Symanowski performed his analysis, the repeated measures test was one of the best methods available for analyzing data collected over time. Thisted Tr. 3059:19-21. Even Zenith’s statistician, Dr. Gibbons, testified that there was nothing wrong with that method, Gibbons Tr. 2190:1-7, and agreed that it showed a statistically significant increase over time in the cholesterol level of the compound ’222-

*864

treated group in comparison to both the olanzapine and control groups. Gibbons Tr. 2214:8-13, 2235:18-2236:3.

260. The repeated measures method was well known to the statistical community in 1990, and Dr. Symanowski learned about it during his graduate training. Symanowski Tr. 680:3-14. Dr. Symanow-ski had previously used the repeated measures method at Lilly before he selected it for use in the D07290 Study. Symanowski Tr. 680:20-681:7. It was an appropriate method for analyzing the D07290 data and was a state-of-the-art method at the time. Thisted Tr. 3059:17-23.

iii) Dunnett’s Test Does Not Show that the Effect of Compound ’222 Is Small

261. Dr. Gibbons performed Dunnett’s test on the data from the D07290 Study and emphasized that at the six-month point he found no statistically significant differences between the high-dose compound ’222 females and the controls. Gibbons Tr. 2192:22-2193:14. He considers those results to be consistent with the Zenith Study, of which he says that “there were at the end of the study very small differences between the control and treated animals.” Gibbons Tr. 2192:22-2193:14.

262. Dr. Gibbons used an “end point analysis,” looking at only the result at the end of the study, whereas Lilly looked individually at every time point, not just the end point, when conducting a Dun-nett’s test. Gibbons Tr. 2227:4-2228:6.

263. The court does not give weight to Dr. Gibbons’ opinion. Dr. Gibbons’ end point analysis ignores all, or virtually all; of the data gathered at earlier time points. Gibbons Tr. 2229:1-5. Thus, his analysis ignored the study’s purpose, as stated in the protocol, which was to determine over the course of a chronic treatment period whether there was a difference in the toxicity profile between olanzapine and compound ’222 in beagle dogs. TX 3439 at ZYP 187 713. In addition, the Dunnett’s test has less statistical power or sensitivity than the repeated measures method, because it looks only at data at a single point in time and ignores information from other points in time. Thisted Tr. 3087:19-3088:17; Gibbons Tr. 2229:19-2230:13. A statistically insensitive test can fail to detect statistical significance where a more powerful test will do so. Gibbons Tr. 2230:21-2231:4. Dr. Symanowski’s repeated measures method, Dr. Thisted’s GEE method, and Dr. Gibbons’ mixed effect regression method, all of which showed statistical significance over time, have greater statistical power than Dun-nett’s test. Thisted Tr. 3088:18-3089:5.

264. Yet, even using the less powerful Dunnett’s test, Dr. Gibbons found statistically significant differences between the compound ’222 and control groups at months two, four, and five. Thisted Tr. 3086:16-3087:7; Gibbons Tr. 2236:25-2238:18; TX 3236 at ZG012979-84.

iv) Dr. Gibbons’ UPL Test Does Not Show that the Effect of Compound ’222 Is Small

265. Dr. Gibbons performed two analy-ses of the Zenith Study data using the upper prediction limit (“UPL”) method and determined that the statistical magnitude of the effect of compound ’222 on the cholesterol levels of the high-dose female dogs was small. Gibbons Tr. 2205:17-2212:5. In the first, he determined the UPL for the pretreatment cholesterol values in the Zenith Study and then compared the measurements at the end of the study to those prediction limits. Gibbons Tr. 2208:11-2209:11. In the second, he computed the UPL at the end of the study and compared them to various reference ranges. Gibbons Tr. 2209:12-2210:6.

266. The court finds Dr. Gibbons’ UPL method is not an appropriate way to mea

*865

sure the cholesterol results in this case. First, the test only looks at extreme values and ignores all other quantitative information. Thisted 3111:23-3114:21. The bar it sets is high- — the upper prediction limit in Dr. Gibbons’ analysis of the Zenith data is more than two standard deviations above the mean. Gibbons Tr. 2249:20-2250:16. The UPL test is a statistically weak way of assessing whether there are differences between groups, because it throws away all information about the cholesterol values other than whether they are above or below the line. Thisted Tr. 3119:4-16. Second, there is no evidence that Dr. Gibbons’ UPL analysis has ever been applied to toxicology testing. Thisted Tr. 3121:21-23. Third, UPL analysis has less statistical power than a repeated measures analysis and other techniques (which show a statistically significant difference), and has even less statistical power than Dunnett’s test. Thisted Tr. 3119:4-19. Finally, Dr. Gibbons’ UPL test ignores the effects of estrus, which is the apparent source of some of the high control and olanzapine values. Gibbons Tr. 2241:3-2242:18; Thisted Tr. 3119:20-24; TX 3236 at ZG 012998.

(j) The Cholesterol Results

267. Dr. John McGrath (“Dr. McGrath”), the senior clinical pathologist on the D07290 Study, Emmerson Tr. 612:11-12, 621:9-11, worked at Lilly as a clinical pathologist from 1979 until his retirement in 2002. McGrath Dep. 6:11-23. He was involved in approximately 50-70 clinical pathology evaluations per year. McGrath Dep. 8:6-23.

268. Dr. McGrath concluded that the most important and biologically significant finding from the D07290 Study was the difference between cholesterol levels measured at 8 mg/kg in the female dogs dosed with compound ’222 compared to those dosed with olanzapine. McGrath Dep. 85:13-89:12. 269. Dr. Emmerson, a toxicologist at Lilly for close to thirty years, reviewed Dr. McGrath’s conclusion in the context of his overall analysis and consideration of the results from the D07290 Study. Dr. Em-merson confirmed that there was a real difference in the two compounds based on the pattern of onset, persistence, and magnitude of the elevation of the effect of compound ’222 on cholesterol in the 8 mg/kg female dogs. Emmerson Tr. 518:15-21, 583:18-22; TX 1001.1 at FH 142.

i) The Increase in Cholesterol Is Biologically Significant

270. Dr. McGrath’s role as a clinical pathologist is to determine, among observed clinical chemistry changes, which of those are of potential biological significance. Rebar Tr. 2308:4-19, 2315:8-2316:15, 2318:15-18 (discussing biological significance); McGrath Dep. 148:14-18 (same).

271. The D07290 Study demonstrated that, of the clinical chemistry changes observed, the increase in cholesterol in the high-dose compound ’222-treated female group was of the greatest potential biological significance because of its early onset, persistence, and magnitude. McGrath Dep. 89:3-12, 124:20-125:9. The high-dose compound ’222-treated female group exhibited a statistically significant cholesterol increase over time compared to the corresponding high-dose olanzapine-treated group and the untreated control group. The comparable olanzapine-treated group did not show a statistically significant increase compared to the control group over the six-month study time period. Emmer-son Tr. 583:18-22. The magnitude of the mean cholesterol corresponded at study termination to a 40-50 mg/dl increase. TX 1001.1 at FH 152; Thisted 3108:9-3109:6;

see also

Findings of Fact ##209, 254.

*866

The effect was evident early in the study and persisted through study termination. McGrath Dep. 89:3-12; 124:20-125:9.

272. The relationship between total cholesterol and heart disease is fairly linear; the greater the cholesterol level, the greater the risk for heart disease. Davidson Tr. 2933:20-22. A drug that raises cholesterol values in humans “should be avoided unless there’s some life threatening condition that would result in a better risk/benefit ratio.” Davidson Tr. 2933:5-7. This is especially important to those living with schizophrenia, who face chronic administration of a drug in which many years of therapy could lead to an increased risk for heart disease. Davidson Tr. 2933:8-15.

273. Many antipsychotic drugs, cloza-pine and olanzapine included, have been noted to be associated with weight gain. It has long been known that some people who gain weight experience an associated rise in cholesterol, and a similar phenomenon has been observed with users of anti-psychotic drugs, including clozapine and olanzapine. The discussion of these findings in the recent literature confirms the importance of avoiding antipsychotic drug candidates like compound ’222 that have the potential to raise cholesterol directly. Paul . Tr. 141:3-142:3, 159:18-160:17; Davidson Tr. 2905:16-2906:6; TX-3883 (“Given the serious implications ... attributable to ... elevated cholesterol, clinicians need to be aware of these risk factors when treating patients with chronic schizophrenia.”).

274. Given these facts, the results of the D07290 Study were not only statistically significant, but also biologically significant.

ii) The Increase in Cholesterol Was Not Caused by Other Factors

a)The Female Estrous Cycle

275. Zenith’s Dr. Nachreiner testified that due to hormonal changes that occur in female dogs during estrus, the cholesterol levels in these dogs naturally rise. Na-chreiner Tr. 1032:19-1033:13; see

also

Gad Tr. 1698:13-16.

276. As mentioned, Lilly did not randomize the dogs involved in the D07290 Study for estrus and did not take the estrous cycle into account when analyzing the data from the D07290 Study. Na-chreiner Tr. 1046:2-5; Finding of Fact # 233.

277. The statistical evidence gleaned from Zenith’s MPI Study confirmed that the cholesterol effect seen with compound ’222 was not attributable to estrus.

See

Findings of Fact ## 209, 254-56.

278. The evidence does not establish that the cholesterol effect seen with compound ’222 is attributable to estrus or that the failure to.take estrus into account confounded the study results..

b) Double Rations

279. Dr. Nachreiner also testified that the double rations fed to Dog 240584 and Dog 242547 — two of the high-dose compound ’222-treated dogs — confounded the D07290 Study results. Nachreiner Tr. 1061:17-1062:8.

280. The evidence reflects that the extra rations fed to those dogs did not affect the cholesterol results from the D07290 Study in a statistically significant manner.

See

Findings of Fact § IV.B.7.e(3)(h)(ii). Therefore, the fact that the dogs were fed double rations did not confound the D07290 Study. This finding is supported by DRL’s Calvert Study. Findings of Fact § IV.B.7.e(2)(c).

c) The Alleged Hypothyroid Dog

281. Hypothyroid dogs commonly have elevated cholesterol levels. Nachreiner Tr. 1072:17-22.

*867

282. In a preclinical examination on July 30, 1990, the clinician at Lilly observed that Dog 239924 had a thin, oily coat and the absence of a patellar reflex, all potential symptoms of hypothyroidism. Nachreiner Tr. 1067:3-10; TX 3454 at ZYP 539 386. This female dog was in the high-dose compound ’222 group.

283. Immediately prior to the beginning of the D07290 Study, Dog 239924 was examined again. The record from that examination indicates that the dog was normal at that time. TX 3454 at ZYP 539 386; Nachreiner Tr. 1143:1-20.

284. Lilly’s Dr. Jeffrey Engelhardt (“Dr. Engelhardt”) personally conducted the histopathology examinations for all of the dogs in the D07290 Study and personally observed the condition of Dog 239924’s thyroid gland. Engelhardt Tr. 2275:5-11, 2277:6-8, 2287:1-3. Dr. Engelhardt noted in the contemporaneous necropsy record and testified at trial that the changes in the dog’s thyroid gland were not pathologically significant and did not warrant any diagnosis of hypothyroidism. Nor was the dog’s coat oily or thin at the time of necropsy. Engelhardt Tr. 2286:6-18. In addition, the symptoms of hypothyroidism are rarely seen in dogs under one year of age, and Dog 239924 was under one year of age at the time of the observed symptoms. Nachreiner Tr. 1142:20-23, 1148:24-1149:9.

285. Zenith’s expert, Dr. Nachreiner, testified that there are many other conditions which may present the same physical symptoms (such as an oily, thin hair coat) as hypothyroidism. Nachreiner Tr. 1143:21-24 (stating that a “laundry list” of other conditions could cause these symptoms), 1144:15-1145:12.

286. Dr. Nachreiner did not perform a necropsy on Dog 239924 and thus never observed the condition of the dog’s thyroid gland.

287. He also admitted that eight tests necessary to diagnose hypothyroidism in the absence of a biopsy were never performed. Nachreiner Tr. 1145:13-19.

288. Although the dog had “minimal lymphocytic thyroiditis,” Dr. Nachreiner admitted that that condition does not mean that the dog had a hypothyroid. Na-chreiner Tr. 1147:23-25; TX 4411 at 486. Indeed, Dr. Engelhardt’s observation of “minimal” lymphocytic inflammation meant only that he saw “barely perceptible changes” leaving more than 90% of the gland “normal and intact.” Engelhardt Tr. 2288:8-18.

289. The evidence is not sufficient to conclude that Dog 239924 suffered from hypothyroidism.

iii) The Reference Range

290. The reference ranges set forth in Trial Exhibit 1166 were used by Lilly at the time they conducted the D07290 Study. Nachreiner Tr. 1023:12-14; TX 1166.

291. The mean cholesterol values for the dogs in the D07290 Study, including those in the high-dose compound ’222 group, were within Lilly’s reference ranges. Nachreiner Tr. 1026:2-8, 1183:5-8.

292. In an experimental study, comparison should be made with the contemporaneous control group in the study. Reference ranges are generally used only when control groups are not available. Rebar Tr. 2366:18-2367:18. In the D07290 and Zenith Studies, cholesterol values from the concurrent controls were reported and available.

293. Reference ranges for any given clinical chemistry parameter are based upon measurements of the parameter in a relatively large group of individuals that are not overtly ill. Thus, the “reference range” represents a range of individual

*868

values that might be seen in a population. Gad Tr. 1812:16-1813:1. It is not a range for mean values, nor is it a range of “healthy” values. Gad Tr. 1816:7-1817:14; Davidson Tr. 2935:25-2936:5.

294. Because the reference range is simply a range of values in a given population, the fact that the mean cholesterol levels of the dogs in the D07290 Study remained within the reference ranges for dog cholesterol does not refute the D07290 Study results.

295. The significance of the finding in the Lilly D07290 Study was that the mean cholesterol value for the whole group increased with compound ’222. If a comparable shift in the mean were seen in the human population, such a shift would-have a significant detrimental effect on public health. Davidson Tr. 2941:5-2943:23; Thisted Tr. 3106:20-3107:5; LD 329. This would be true even if that increased mean value were numerically still within the reference range. Rebar Tr. 2320:10-24; Thisted Tr. 3106:6-3107:10.

iv) Other Statistically Significant Changes

296. In addition to the cholesterol finding, Dr. Symánowski’s analysis revealed several other statistically significant differences between groups treated with compound ’222 and olanzapine. For example, compound ’222 slightly increased alanine aminotransferase, gammaglutamyl transf-erase, and total protein, whereas olanza-pine slightly increased several erythrocytic parameters, bilirubin, and albumin. Rebar Tr. 2335:1-2347:21; LD 121; LD 123; LD 124; LD 125; Pentel Tr. 1890:23-1891:4; TX 1171 at ZYP 427 155-56 (showing statistical significance between olanzapine and compound ’222 for numerous parameters including ERYS (erythrocytes), HGB (hemoglobin), PCV (packed cell volume), and T BILL (bilirubin)); Symanowski Tr. 723:13-727:13; TX 3451 at ZYP 427 135-37.

297. The determination of whether or not a statistically significant change is clinically significant depends' on context. McGrath Dep. 121:2-122:22; Rebar Tr. 2335:8-24.

298. The albumin levels were in the 3.3-3.5 g/dl range. Rebar Tr. 2346:23-2347:3; McGrath Dep. 172:17-21. A clinical pathologist would only be concerned with these levels if they reached 5 g/dl. McGrath Dep. 172:22-24. Thus, the statistically significant increases in albumin in the male dogs dosed with 8 mg/kg of olan-zapine were not biologically significant. Rebar Tr. 2347:4-21; LD 121.

299. The total bilirubin values were in the 0.1-0.2 mg/dl range, but a clinical pathologist would not be concerned until those values were closer to 1.0-2.0 mg/dl. McGrath Dep. 96:3-22; Rebar' Tr. 2346:3-12. Thus, although statistically significant, the increases in bilirubin in the male dogs dosed with 8 mg/kg of olanzapine were not biologically significant. McGrath Dep. 96:13-22; Rebar Tr. 2344:5-12; LD 122.

300. The erythrocytic parameters comprising hemoglobin, erythrocyte count, and packed cell volume were also found to be elevated in a statistically significant manner in dogs, but these elevations were not biologically significant. Rebar Tr. 2335:25-2344:2; LD 123; LD 124; LD 125. These three parameters are a measurement of the red blood cell mass in the blood stream, and they usually increase or decrease together, as was the case in the D07290 Study. Rebar Tr. 2336:3-2337:24. Based upon the trends in the data, the olanzapine and control dogs’ erythrocytic parameters increased, while the compound ’222 dogs’ erythrocytic parameters decreased. Rebar Tr. 2338:3-22. While this difference in values between the olan-zapine and compound ’222 dogs may be

*869

statistically significant, the magnitude of the change was not biologically significant. Rebar Tr. 2338:23-2339:4, 2343:8-24.

301. Of all the changes observed differentiating the compounds, the difference in cholesterol was the most striking difference between olanzapine and compound ’222. McGrath Dep. 143:19-144:3, 149:16-22.

v) Dog Data Excluded from the Study Did Not Effect the Study Findings

302. The removal of data from Dog 240712 and Dog 240692, two female dogs in the high-dose olanzapine group, did not confound the statistical analysis of the D07290 Study.

See

Findings of Fact § IV.E.8.

f. Prolactin as a Previously Unconsidered Unexpected Result

(1) The Results of the MPI Study

303. Lilly claims that Zenith’s MPI Study showed that compound ’222 elevates prolactin whereas olanzapine does not.

304. Prolactin is a hormone that is released from the pituitary gland late in the gestation period to promote lactation in animals. Nachreiner Tr. 1009:8-21.

305. Zenith’s study director recommended measuring prolactin after observing mammary enlargement and lactation in the compound ’222-treated dogs. Na-chreiner Tr. 1121:22-1122:19. Zenith’s protocol was then modified to measure prolactin at months three, four, five, and six, Nachreiner Tr. 1125:21-25, when contemporaneous samples from the control and other groups would be available for comparison.

306. There were no pretest prolactin levels taken for any of the dogs. Similarly, there were no prolactin levels taken for any of the dogs in week one of the study, at month one of the study, or at month two of the study. Concannon Tr. 2601:4-14.

307. Prolactin measurements were taken from the forty-four dogs in the study at month three. Concannon Tr. 2601:16-20. The mean prolactin level in the dogs given compound ’222 was statistically significantly higher than the olanzapine or control groups. Nachreiner Tr. 1119:11-1120:9; LD 37.

308. After Zenith obtained those results, the protocol for the study was amended to cancel the rest of the prolactin measurements. Nachreiner Tr. 1126:19-22. Samples for measuring prolactin were taken at month four but never measured. Nachreiner Tr. 1127:11-1128:13.

309. A prolactin measurement that is taken at a single point merely indicates whether a dog is releasing prolactin at that particular point in time, but it does not indicate anything of biological significance. In order to obtain biological significance, there must be measurements taken at the commencement of the study to establish a baseline level, as well as measurements regarding prolactin release both prior to and after the particular measurement at issue. Nachreiner Tr. 1009:22-1010:7. This did not occur. Thus, the fact that there was a statistically significant difference between the prolactin levels of the dogs given compound ’222 and those given olanzapine during month three of the MPI Study, Nachreiner Tr. 1119:11-1120:9, provides insufficient information upon which to make a comparison between the prolac-tin levels of compound ’222-treated dogs and those of the olanzapine-treated dogs. Nachreiner Tr. 1013:12-19;

see also

Gad Tr. 1727:22-1728:1,1728:22-1729:14.

310. The other observations made by the researchers working on the MPI Study, including excessive mammary development and lactation, Concannon Tr. 2559:2-11, may support the theory of increased prolactin with treatment, but there are no measurements beyond this single

*870

point in time to confirm this theory or to indicate its significance.

See

Nachreiner Tr. 1009:22-1010:7; Gad Tr. 1727:22-1728:1.

311. The results of the MPI Study are not sufficiently complete and, as a result, not sufficiently reliable, to support any conclusions regarding the effect of compound ’222 on prolactin.

(2) The Results of the Calvert Study

312. With regard to DRL’s Calvert Study, there is no evidence that any pro-lactin measurements were recorded.

313. Dr. Patrick Concannon (“Dr. Con-cannon”), Lilly’s expert on mammalian reproduction, endocrinology, and physiology, Concannon Tr. 2553:10-22, saw 239 recorded instances of mammary swelling in the compound ’222-treated dogs and 195.re-corded instances of mammary swelling in the olanzapine-treated dogs. Concannon Tr. 2635:12-21. In contrast, there were only four recorded instances of mammary swelling in the control dogs. Concannon Tr. 2635:22-2636:1.

314. Dr. Concannon saw sixty-one recorded instances of lactation in the olanza-pine-treated dogs. Concannon'Tr. 2637:1-11. He also observed that the incidence and duration of lactation was greater in the compound ’222-treated dogs as compared to the olanzapine-treated dogs. Concannon Tr. 2637:10-2638:15. In the control group, there were no instances of lactation. Concannon Tr. 2637:12-14.

315. The fact that the researchers recorded some lactation and mammary swelling in the dogs during the Calvert Study does not establish a prolactin-related unexpected difference between olanzapine and compound ’222.

(3) Olanzapine’s Effect on Prolactin

316. At the time olanzapine was approved by the FDA, its labels stated, and continue to state today, that olanzapine elevates prolactin levels and that a modest elevation persists during chronic administration of the drug. Kinon Tr. 2505:13-22, 2522:5-12, 2525:25-2526:4; TX 3877.

317. When olanzapine was approved, the FDA asked Lilly to include boilerplate prolactin language that was utilized with all first-line antipsychotic agents at that time. Kinon Tr. 2505:13-22.

318. Dr. Bruce Kinon (“Dr. Kinon”) is employed by Lilly as a medical advisor and was certified as an expert on the effects of antipsychotic drugs on prolactin. Kinon Tr. 2493:20-23, 2498:7-9. Dr. Kinon testified that since olanzapine’s approval, extensive research has confirmed that olan-zapine does not elevate prolactin in most patients. Kinon Tr. 2505:25-2506:8. Dr. Schulz agreed that olanzapine does not increase prolactin in patients in any significant way, Schulz Tr. 2978:8-12. He based his opinion on review of the scientific literature in refereed journals, such as the Journal of Clinical Psychiatry. Schulz Tr. 2980:1-5, 2982:17-24.

319. The olanzapine label still reflects a concern about increased prolactin. TX 3877.

320. Changing a drug’s label is a long, expensive, and uncertain process because it is ultimately up to the FDA to determine whether the label will be changed. Schulz Tr. 2983:21-2984:5.

321. The totality of the evidence does not establish that there are prolactin-relat-ed unexpected differences between compound ’222 and olanzapine.

g. Unexpected Differences Between Olanzapine and Flumezapine

322. Lilly asserted to the PTO that olanzapine unexpectedly failed to abnormally elevate liver (SGOT and SGPT) and muscle (CPK) enzymes in human patients compared to flumezapine. TX 1001.1 at

*871

FH 14-15, 17, 45. This claim went unchallenged by the PTO. Zenith now challenges that finding.

(1) Liver and Muscle Enzyme Test Results from the Flumezapine Clinical Trials

323. Lilly first tested flumezapine (compound LY120363) in healthy human volunteers from January through October 1981. TX 3737.

324. Dr. Pullar, a Lilly pharmacologist in the U.K., was chair of the flumezapine project team. Pullar Tr. 166:5-6, 185:20— 24. Dr. Lewis Lemberger (“Dr. Lember-ger”), a clinical pharmacologist at the Lilly Laboratory for Clinical Research (“Lilly Clinic”) in Indianapolis, was a member of the flumezapine project team. Lemberger Tr. 2449:8-12.

325. Following the second Phase I clinical trials with flumezapine in healthy human volunteers, Dr. Lemberger concluded that flumezapine — in the doses he administered — did not adversely affect the liver enzymes of those healthy human volunteers in a clinically significant manner. Lemberger Tr. 2454:10-15.

326. In the spring of 1982, Lilly commenced its Phase II clinical trials of flume-zapine in actual schizophrenic patients at various test sites in the United States, England, and Scotland. TX 1015, TX 3739 at ZY 1762 1383-84. Dr. David Dunner of Seattle, Washington, enrolled one patient; Dr. Jay Cohn of Newport Beach, California, enrolled six patients; Dr. Lawrence Gosenfeld (“Dr. Gosenfeld”) of Los Ange-les, California, enrolled six patients; Dr. S.A. Montgomery (“Dr. Montgomery”) of London, England, enrolled three patients; and Dr. Iain Glen of Inverness, Scotland, enrolled one patient. TX 3739 at ZYP 1383-84; Diamond Tr. 1330:13-17.

327. The data from those five test sites revealed that the enzyme levels in those patients administered flumezapine increased significantly.

See

TX 3739. In particular, four of the six subjects at Dr. Gosenfeld’s site experienced CPK (muscle enzyme) elevations that were significant enough to be a concern. Diamond Tr. 1353:10-23; TX 3739 at ZY 1762 1386-1762 1402. In an assay that specified the upper range of normal to be 280, four flumeza-pine patients (041, 042, 043, and 044) experienced CPK increases up to 5504, 5000, 6352, and 3090, respectively. TX 3739 at ZYP 1762 1393-96.

328. In Lilly’s report, Lilly stated that “[t]he elevations of CPK are not well explained ... Dr. Gosenfeld has reported that, both before, during and after this study elevated CPKs were seen in patients not in the study ...” TX 1015 at ZYP 177 1999; TX 3739 at ZY 1762 1385.

329. Lilly reported these results to the FDA in April 1982. The FDA advised Lilly to terminate the Phase II flumeza-pine clinical trials. Pullar Tr. 188:21-23; TX 3259; TX3260.

330. In December 1985, a joint U.S./ U.K. team was set up to find a replacement for flumezapine. TX 1058 at ZYP 520 531.

(2) Liver and Muscle Enzyme Test Results from the Olanzapine Clinical Trials

331. Dr. Lemberger and Dr. Mark Goldberg (“Dr. Goldberg”) conducted three Phase I clinical trials of olanzapine in normal human volunteers at the Lilly Clinic in Indianapolis. Goldberg Tr. 355:18-356:9; Lemberger Tr. 2456:8-10; 2456:18-21. The Phase I clinical trials included HGAA in September 1986, HGAB in February 1987, and HGAC in May 1987. Goldberg Tr. 307:9-308:3; TX 3741 at ZYP 289 1095-96; TX 3742 at ZYP 289 1449.

332. In the interim report for the Phase I clinical trials of olanzapine, Dr.

*872

Lemberger reported that five out of eleven patients had elevations in their SGPT and/or SGOT. The doses administered to these volunteers were five to ten times less than those projected to be administered to actual schizophrenic patients. TX 3745 at ZYP 506 148-49; Lemberger Tr. 2466:24-2469:7.

333. In the final report for the olanza-pine Phase I clinical trials, Dr. Lemberger stated, “Two of the three subjects completing this study experienced significant elevations in liver enzymes. Based on these results in normal human volunteers, care should be taken to cautiously administer LY170053 [olanzapine] in patients with schizophrenia, with special attention given to the liver enzyme changes that may occur.” TX 3744 at ZYP 289 1617; see

also

Goldberg Tr. 361:2-5.

334. Following the’ conclusion of the tests in healthy human volunteers, Lilly formulated a clinical plan to test olanza-pine in actual schizophrenic patients. TX 3688.

335. The study, designated as E001, began in the U.K. in December 1988, -and was completed in July 1990.

See

TX 1058; TX 1064.

336. Dr. David Wheadon (“Dr. Whea-don”), the clinical research physician on the olanzapine project team, reported in April 1990 that both flumezapine and olan-zapine had the “propensity to cause elevations in liver enzymes [SGOT and SGPT].” He reported further:

In the case of flumezapine, these elevations were seen in schizophrenic patients at doses generally higher than 15 mg q.d.... Both normal volunteers and patients exposed to LY170053 have, in some cases, experienced elevations in liver enzyme levels. Of the elevations in normal volunteers, one subject experienced this event at a dose of 2.5 mg, the other cases were at doses of 6 mg or higher. In schizophrenic patients, LY170053 was associated with enzyme elevations at a 30 mg dose in one case and a 10 mg dose in two others.

The propensity to cause elevations in CPK appears to be more of a discriminating parameter when comparing these two compounds than elevations in liver enzymes. Admittedly one can argue that had LY170053 been dosed as high as the doses given flumezapine patients, that similar results may have been seen with LY170053. In addition, the literature on elevations of CPK in psychotic patients has featured an ongoing discussion concerning the etiology of this phenomenon, with some researchers viewing it as a frequently accompanying symptom of psychosis and others noting this elevation to be a neuroleptic side effect.

TX 1063 at ZYP 449 1142.

(3) Dr. Diamond’s Opinion

337. Zenith hired Dr. Ronald Diamond (“Dr. Diamond”), a clinical psychiatrist and psychopharmacologist, as an expert for purposes of this litigation. Diamond Tr. 1314:4-13, 1315:1-2, 1325:21-1326:2. Dr. Diamond reviewed Lilly’s study data from the flumezapine and olanzapine clinical trials, admitted into evidence as TX 3739, TX 3741, TX 3744, and TX 3747. Diamond Tr. 1326:3-1327:10. He testified that in his opinion, there is no difference between the profile of flumezapine and olanzapine based on liver enzymes. Diamond Tr. 1327:24-1328:9, 1345:22-1346:1.

338. Dr. Diamond reported that with regard to the flumezapine clinical trial, in which schizophrenic patients' were given between 0 and 40 mg/day,

see

TX 3739 at ZY 1762 1393 (Patient 41), one group of patients experienced a spike in the elevation of liver enzymes, and the elevation came back to normal after the drug was

*873

withdrawn. Diamond Tr. 1347:7-14. Another group of patients experienced a spike during the administration of the drug, but that elevation appeared to return to near normal before the drug was withdrawn. Diamond Tr. 1348:1-16.

339. Dr. Diamond observed these same liver enzyme patterns in the olanzapine clinical trials. Diamond Tr. 1349:4-8, 1350:10-11; TX 3741; TX 3744; TX 3747.

340. Of the fifteen olanzapine subjects reviewed by Dr. Diamond — some normal volunteers and some schizophrenic patients — five had liver enzyme elevations. Diamond Tr. 1351:15-18; TX 3741; TX 3744; TX 3747. Of the fifteen flumezapine subjects reviewed by Dr. Diamond, six had some degree of abnormal liver enzyme elevation. Diamond Tr. 1351:10-14; TX 3739.

341. Dr. Diamond also testified that in his opinion, “there was no reason to believe that there were clinically significant risks from differences in CPK data that was presented in the data.” Diamond Tr. 1328:10-12.

342. In support of his opinion, he testified that the elevated CPK levels were all from Dr. Gosenfeld’s test site. Diamond Tr. 1390:12-1391:12.

343. Of the four subjects experiencing CPK elevations at the Gosenfeld test site, one of the subjects had elevated CPK levels before taking flumezapine. Diamond Tr. 1354:7-24. This patient began the study with a CPK level of 681 mU/ml, but after 19 days of flumezapine administration, his CPK increased to 5504 mU/ml. TX 3739 at ZY 1762 1393.

344. Also, in all four subjects who experienced elevated CPK, the CPK levels either came down or even normalized while the subjects continued to take flume-zapine. Diamond Tr. 1354:11-15, 1387:20-1389:2.

See e.g.,

TX 3739 at ZY 1762 1393 (Patient 41).

345. The two other test subjec

This text is long and has been trimmed here. Open the source document for the complete record.