OPINION

Table of Contents

I. The Legal Backdrop and. the Positions of the Parties.401

II. Proceedings on the Motion.402

*400 III. Scientific Background.

A. Relevant Physiology.

1. Cells .

2. The Liver.

3. Mechanisms of Cell Death.

B. Drugs, Toxicity to the Liver, and Rezulin

C. Patient Population.

D. Evidence of Causation in Medicine.

IV.The Proposed Testimony.

A. The Experts and Their Opinions. Lr*

1. Dr. Smith. Ir**

2. Dr. Reed.

3. Dr. Julie.

4. Dr. Day.

5. Dr. Bonkovsky.

6. Summary.

B. The Science upon Which the Experts Rely. H

1. Early Links in the Proposed Causal Chain: The Claim that Rezulin Causes Apoptosis Through Effects on the Mitochondria or the BSEP-..'..

a. Studies Connecting Rezulin to Apoptosis.

b.. Studies Connecting Rezulin to Mitochondrial Damage.

c. Studies Connecting Rezulin to an Effect on the BSEP.

2.' The Last Link in the Chain: The Claim that Apoptosis from Rezulin ' Causes Silent Injury....'

C. Arguments that Rezulin Can Cause Silent Liver Injury Through Mechanisms Other than Apoptosis.:..

D. Patients Whose Liver Enzymes Were Not Monitored.

V.Law Governing the Admission of Expert Testimony ..

A. Daubert and Its Progeny.

B. The Daubert Standards Apply to Opinions About General and Specific Causation.

VI. Daubert Analysis of the Proposed Testimony.423

A. Testing and Error Rate, Peer-Review, Publication, Widespread Acceptance...'..423

B. Independence from Litigation....424

C., Consideration of Contrary Evidence.426

D. “Fit” and the “Analytical Gap” .426

1. The Experts Have No Evidence for the Crucial Link in Their Causal Chain. 426

2. The. Experts Have Failed To Link the Studies on Mitochondria and the BSEP into Their Causal Chain.427

3. The Research on Apoptosis in Cell Cultures Does Not “Fit” the Opinion at Issue ..428

a. Types of Cells ..-.429

b. Dose.430

VII. The Plaintiffs’ Argument Concerning Differential Diagnosis.435

VIII. Conclusion.437

In 1996, Warner-Lambert Company (“Warner-Lambert”), now owned by Pfizer íñe., announced the development of Rezu-lin, the trade name of a drug used to treat Type 2 diabetes, a disease affecting ap *401 proximately 16 million Americans. 1 The United States Food and Drug Administration approved the drug in 1997, and it was administered to 1.92 million people. Following reports that some patients taking Rezulin experienced liver failure resulting in transplant or death, the drug was withdrawn from the market in March 2000. 2 This led to the commencement of thousands of lawsuits for alleged personal injuries or apprehension of personal injuries in consequence of the ingestion of the drug. 3 The federal actions have been consolidated in this Court for pretrial proceedings. 4

Extensive liability discovery against defendants has been completed. Defendants Pfizer Inc., Warner-Lambert, and Parke-Davis 5 (collectively, “Pfizer”) move, pursuant to Federal Rule of Evidence 702 and Daubert v. Merrell Dow Pharmaceuticals, Inc., 6 to exclude “proposed expert testimony that Rezulin can cause a liver injury, or exacerbate a pre-existing liver condition, in the absence of marked elevation of liver enzymes while the patient was taking the medication.” 7 The Plaintiffs’ Executive Committee, which is responsible for coordinating the activities of plaintiffs during pretrial proceedings, 8 retained the experts and is defending this motion.

I. The Legal Backdrop and the Positions of the Parties

Causation in toxic tort cases has two components, general and specific, and the plaintiff must establish both in order to *402 prevail. 9 “General causation is whether a substance is,capable of causing a particular injury or condition in the general population, while specific causation is whether a substance caused a particular individual’s injury.” 10 As explained in the Federal Judicial Center’s Reference Manual on Scientific Evidence:

“General causation is established by demonstrating, often through a review of scientific and medical literature, that exposure to a substance can cause a particular disease (e.g., that smoking cigarettes can cause lung cancer). Specific, or individual, causation, however, is established by demonstrating that a given exposure is the cause of an individual’s disease (e.g., that a specific plaintiffs lung cancer was caused by his smoking).” 11

Plaintiffs offer the testimony of a number of expert witnesses to prove general causátion. As relevant here, they would opine that Rezulin is capable of causing liver injury “silently,” that is without markedly elevating liver enzymes, and that such injury is a consequence of a form of liver cell death, known as apoptosis, that the experts assert can be induced by Rezu-lin.

Pfizer contends that plaintiffs’ experts’ testimony is insufficiently reliable to satisfy Daubert and Rule 702. It claims that their theories are unsupported by testing and that the potential rate of error therefore cannot be determined. They maintain also that they have not been subjected to peer review and publication and that they do not have widespread acceptance in the scientific community. Furthermore, Pfizer argues that the opinions of plaintiffs’ experts are not the product of independent research but were developed solely for this litigation and, finally, that the experts overlook or ignore contrary evidence. The plaintiffs resist these assertions.

11. Proceedings on the Motion

Each side has submitted thousands of pages of expert reports, scientific and medical articles, depositions and other documents in connection with the motion. The Court heard oral argument and then conducted a two-day evidentiary hearing at which three of the plaintiffs’ experts and one defense expert testified. 12 The Court then had the benefit of extensive proposed findings of fact from each side and point-by-point responses .to each set of proposed findings.

III. Scientific Background

A. Relevant Physiology

1. Cells

The human body consists of cells. The cell has three basic components: the nucleus, the cytoplasm, and the plasma mem *403 brane. 13 The nucleus contains the cell’s genetic information. 14 The plasma membrane surrounds the cell and “acts as a selective barrier that enables the cell to concentrate nutrients gathered from its environment and retain the products it synthesizes for its own use, while excreting waste products.” 15 The cytoplasm is defined as everything besides the nucleus and cell membrane, 16 and it contains among other things membrane-bound structures known as organelles, of which one type is the mitochondrion (in the plural, mitochondria). 17 The mitochondria sometimes are referred to as the “powerhouse” of the cell 18 because chemical reactions occurring therein produce adenosine triphosphate (“ATP”), the cell’s energy currency. 19

2. The Liver•

The liver, located in the right upper quadrant of the abdomen, is a large organ that plays a central role in the body’s biochemical activity. It metabolizes, or breaks down, a large number of substances, synthesizes essential enzymes, and detoxifies and eliminates a variety of compounds from both within and without the liver. 20 A vein known as the portal vein carries to the liver all blood from the digestive tract, facilitating the liver’s absorption of toxic chemicals that a person has ingested. 21 Most of the liver’s. biochemical functions take place within cells known as hepatocytes, which are the predominant type of cell in the liver. 22

Medicine recognizes many forms of liver disease, including non-alcoholic fatty liver disease (“NAFLD”), described as “fatty liver without inflammation,” 23 and a related condition 24 known as non-alcoholic stea-tohepatitis (“NASH”), described as “fatty liver with inflammation.” 25 These conditions are associated with fibrosis, 26 de *404 scribed as “scarring of the liver” 27 and more precisely as “the presence of excess extracellular matrix.” 28 Fibrosis can lead to cirrhosis, 29 a final stage of liver disease in which “groups of hepatocytes become completely encased by fibrous material, and nodules replace normal lobular organization.” 30 When liver disease is “moderately advanced,” “fibrosis often appears as connective tissue that spans portal and central areas,” a condition known as “bridging fibrosis.” 31

3. Mechanisms of Cell Death

Some of the plaintiffs’ witnesses base their proposed testimony on theories relating to death of liver cells allegedly attributable to Rezulin.

There are two basic mechanisms of cell death: necrosis and apoptosis. 32 Experts from both sides agree on their basic features.

In necrotic cell death, the cell membrane ruptures, and the contents of the cell are discharged into the surrounding tissue. Among these cellular contents in the liver are enzymes, including alanine amino-transferase (“ALT”) and aspartate amino-transferase (“AST”) (collectively, “transa-minases”). 33 Thus, necrotic cell death in the liver in sufficiently large quantities is accompanied by elevations of liver enzyme concentration in the blood 34

In apoptosis, by contrast, the dying cell shrinks and is engulfed and digested by neighboring phagocytes, 35 described in one of the expert reports as “tissue clean up cells.” 36 Apoptosis sometimes is known as “programmed cell death” or “cell suicide.” 37 Like necrosis, it is “often naturally initiated by the body to eliminate cells that are no longer needed” 38 and is necessary for human survival. 39

Apoptosis in the liver usually is not accompanied by enzyme elevation because the phagocytes typically devour the apop-totic cells before the latter release their contents into blood. Apoptosis in the liver, however, may result in elevated liver enzymes when it takes place on such a scale that the ability of viable cells to remove dying cells is overwhelmed. 40

*405 B. Drugs, Toxicity to the Liver, and Re-zulin

Because the liver “is central to the metabolic disposition of virtually all drugs and foreign substances,” liver injury “is a potential complication of nearly every medication that is prescribed.” 41 Experts from both sides agree that liver injury caused by drugs is typically categorized as hepa-tocellular, cholestatic, or mixed based on the pattern of blood test results. Hepato-cellular injury is injury to hepatocytes and is manifested by elevated quantities of transaminases in the blood. Cholestatic injury is injury to the liver’s ability to produce and excrete bile and is manifested by elevated quantities of alkaline phosphatase and bilirubin. The injury is mixed if features of both hepatocellular and choles-tatic injury are present. 42

In general, drugs are either predictable or unpredictable (idiosyncratic) toxins to the liver, or hepatotoxins. Predictable toxins are dose-dependent. Unpredictable toxins, while capable of producing damage at therapeutic doses, do. so rarely. 43

There is no dispute that a small percentage of patients treated with Rezulin developed reversible elevations of ALT of more than three times the upper limit of normal. 44 The mechanism by which Rezulin is toxic to liver cells is not well understood. 45 The defendants’ experts believe that Rezu-lin is an idiosyncratic hepatotoxin, inducing liver injury on an unpredictable basis in a very small number of cases. 46 The plaintiffs’ experts believe that Rezulin can be directly toxic to cells, even though the drug exhibits features of idiosyncratic toxicity. 47 Liver injury attributable to Rezu-lin is predominantly hepatocellular, 48 but the defendants do not dispute, at least for purposes of this motion, that the drug is capable of causing cholestatic or mixed injury. 49

*406 C. Patient Population

Type 2 diabetes affects an estimated 16 million Americans. 50 It is associated with obesity, and both obesity and diabetes are associated with NAFLD and NASH. 51 Approximately 50 percent of patients with diabetes have NAFLD. 52 (The prevalence of NAFLD in the general population of the United States was once estimated at 24 percent and is likely higher today. 53 ) The percentage of diabetics with NASH is unknown but probably smaller. 54

D. Evidence of Causation in Medicine

The expert testimony in this ease purports to be grounded on a number of scientific and medical studies. For the sake of context, the Court briefly reviews principles regarding the applicability of different kinds of scientific and medical research.

The “gold standard” for determining the relationship between a drug and a health outcome is the clinical trial. In such investigations, subjects are assigned randomly to one of two groups: one exposed, and the other not exposed, to the drug of interest. Such studies, however, are not always available, at least in part because ethical constraints preclude exposing human beings to agents known or thought to be toxic. 55 Thus, other kinds of evidence often are used to assess whether a drug or agent is related to the risk of developing a certain condition.

Chief among these are observational epidemiological studies, in which subjects that were exposed to an agent prior to the investigation are compared with subjects not so exposed. 56

Another type of information used by physicians and medical researchers is the case report, which is a description of a particular patient’s clinical history and symptoms. As explained in the Reference Manual on Scientific Evidence:

“Case reports lack controls and thus do not provide as much information as controlled epidemiological studies do. However, case reports are often all that is available on a particular subject.... Causal attribution based on case studies must be regarded with caution. However, such studies may be carefully considered in light of other information available, including toxicological data.” 57

The difficulty with case reports, in other words, is distinguishing between association and causation. Simply because a patient exposed to a particular substance exhibited a set of symptoms does not mean that it was the substance that caused the symptoms.

In addition to research on humans, scientists often perform experiments on living animals, such as rats, mice, and monkeys. The advantages of such studies include the fact that they can be *407 conducted as true experiments, with exposure controlled and measured and ethical limitations diminished. 58 Nevertheless, although it is “a basic principle of toxicology and pharmacology” that “a compound causing an effect in one mammalian species [usually] will cause it in another,” 59 extrapolation from animal studies to humans cannot be done uncritically. For one thing, different species have important physiological differences. For another, the high doses often used in animal studies may not correspond to considerably lower concentrations of a drug or other substance to which humans are in reality exposed. 60

Finally, researchers frequently conduct experiments on cell and tissue cultures. These experiments, sometimes referred to as in vitro studies to distinguish them from studies performed in vivo, meaning on live humans and animals, also are subject to the problem of extrapolation. It is not always clear that “one can generalize findings from the artificial setting of tissues in laboratories to whole human beings.” 61

TV. The Proposed Testimony

The defendants do not challenge the admissibility of expert testimony to the effect that Rezulin is capable of causing liver injury that results in elevation of liver enzymes. They train their fire only on opinion testimony that the drug is capable of doing so “silently,” that is, without markedly elevating those enzymes approximately concurrently with treatment. 62 The defendants challenge aspects of the opinions of five expert witnesses.

Before reviewing the challenged testimony, a preliminary note is in order. The plaintiffs’ counsel at every stage have gone further than their experts. They have formed hypotheses that the experts nowhere mentioned, they have forged connections that the experts stopped short of drawing, and they have portrayed as sturdy hypotheses that the experts espoused hesitantly if at all. None of these elaborations by counsel is relevant. The subject of this motion is the proposed testimony of expert's, not the theories of the lawyers.

A. The Experts and Their Opinions

1. Dr. Smith

The opinion of Dr. Martyn T. Smith, who testified at the evidentiary hearing, is central to this motion. Dr. Smith is a toxicologist. He holds a Ph.D. and is a professor in the Division of Environmental Health Sciences at the University of California at Berkeley’s School of Public *408 Health. He is not a clinician and has no training in hepatology. 63

Dr. Smith opines that “[a] careful review of all of the literature allows one to conclude to a reasonable degree of scientific probability that [Rezulin] more likely than not ... can damage the liver ‘silently’ because apoptosis does not cause elevation of liver enzymes in serum.” 64 He states “that to a reasonable degree of scientific probability troglitazone 65 can induce apop-totic death of cells in the liver in humans in vivo. ” 66

Dr. Smith offers two major mechanisms by which Rezulin could cause apoptosis. The first involves a cascade of chemical events in which the mitochondria cease functioning normally and release a protein into the cytoplasm that triggers apopto-sis. 67 The second involves the bile salt export pump, or BSEP. The BSEP is a protein in the plasma membranes of hepa-tocytes that pumps bile salt out of the cell. Dr. Smith states that Rezulin can interfere with the functioning of the pump, leading to the build-up of toxic bile salts in the cells and that this condition triggers apop-tosis. 68

2. Dr. Reed

Dr. John C. Reed’s opinion also is a major subject of this motion. Dr. Reed is the president and chief executive officer of the Burnham Institute, a research center in La Jolla, California. 69 He holds an M.D. and a Ph.D. 70 and has described himself as follows: “I don’t consider myself a liver expert or hepatology expert. I consider myself an apoptosis expert.” 71 Dr. Reed testified at the evidentiary hearing.

Dr. Reed’s declarations are intended to “address[ ] the evidence that Rezulin (tro-glitazone) induces apoptosis in human and animal cells.” 72 Dr. Reed’s declarations do not say that a Rezulin-induced injury could be silent. 73 His testimony therefore is relevant only to the extent that it provides support for the other experts — principally Dr. Smith — who are willing to draw that conclusion.

S. Dr. Julie

Neil Julie, M.D., practices gastroenterol-ogy and hepatology in a Maryland suburb of Washington, D.C., 74 and also testified at the evidentiary hearing. His initial declaration is organized for the most part as a set of conclusions, each in the form of a single sentence and each followed by the statement that “[m]y opinions are based upon my education, training and experience as well as my review of the following *409 materials” and then a long list of documents, medical articles, and testimony. The conclusions relevant here are that “Rezulin causes a multitude of hepatocellu-lar injuries including but not limited to ... apoptosis .... ” and “Rezulin liver injury can occur without an elevation of AST and ALT.” 75

The plaintiffs submitted a second declaration from Dr. Julie to address:

“the tools a medical doctor should appropriately utilize in order to diagnose a medical condition generally and, more specifically, the use of differential diagnosis [ 76 ] with respect to a Rezulin-relat-ed hepatic injury, including evidence of studies showing mitochondrial damage and apoptosis caused by Rezulin.” 77

Dr. Julie concludes: “it is my opinion that it would be inappropriate to broadly conclude that a patient did not suffer a drug-induced liver injury from Rezulin simply based upon the absence of a marked elevation in liver enzymes.” 78 In support of this statement, Dr. Julie states among other things that “[a] physician performing a differential diagnosis must consider. ... relevant scientific evidence” of Rezulin’s “toxic effect on mitochondria and its ability to induce cell death via apoptosis.” 79

Dr. Julie acknowledged that his view that Rezulin is toxic to the mitochondria is derived from the opinions of Drs. Smith and Reed and the articles they cite for support. 80

A Dr. Day

Dr. Christopher P. Day is a research scientist and practicing hepatologist. He holds an M.D. and a Ph.D. and is a professor at the University of Newcastle upon Tyne in the United Kingdom. 81 His report is concerned mainly with illustrating the varied ways in which Rezulin and other drugs can injure the liver.

As relevant here, Dr. Day would testify that “a biologically plausible explanation for why many patients with liver disease post-exposure to troglitazone may not have had significant elevations of their transa-minase levels” is that cells undergoing apoptosis are consumed by surrounding cells. 82 Dr. Day’s declaration further states that cholestasis and mitochondrial injury are “biologically plausible mechanisms for the development of apoptosis in patients exposed to troglitazone.” 83

The Court notes that cholestatic injury normally results in increased bilirubin or alkaline phosphatase 84 and that Dr. Day does not say that apoptosis related to cho-lestasis, if it occurs at all, would not affect bilirubin and alkaline phosphatase levels. Furthermore, Dr. Day does not say that *410 apoptosis related to mitochondrial injury could cause injury without an elevation of transaminases. In other words, Dr. Day does not appear to offer the testimony that the defendants are seeking to exclude, which is the opinion that Rezulin can cause liver injury without an elevation in at least one of the four substances. His testimony therefore is of limited relevance. The Court nonetheless considers it insofar as it could be read to suggest that any injury caused by Rezulin would be silent.

5. Dr. Bonkovsky

Herbert L. Bonkovsky, M.D., is a hepa-tologist and a professor at the University of Connecticut Medical School. 85 The thrust of his opinion is that Rezulin “causes a wide spectrum of liver disease.” 86 He states among other things that liver inflammation “such as that caused by Rezulin” can lead to bridging fibrosis and cirrhosis and that “[tjhese injuries have been observed in many cases with either normal or near normal values of ALT.” 87 Dr. Bonkovsky, however, does not say that any such injuries that have been observed were due to Rezulin, nor does he provide any support for the statement that those injuries occurred with normal values of ALT.

Elsewhere in his declaration Dr. Bon-kovsky opines that apoptosis 88 and inhibition of the BSEP 89 are plausible mechanisms of injury due to Rezulin. In addition, Dr. Bonkovsky’s opinion reviews some of the research cited in the reports of Drs. Reed and Smith for the proposition that Rezulin can affect the mitochondria. 90 But Dr. Bonkovsky does not attempt to relate these statements, nor does he state that an effect on the mitochondria could cause apoptosis or that any injuries due to inhibition of the BSEP or apoptosis would be silent. He states without significant elaboration that he agrees with the views expressed in Dr. Smith’s report “on the various mechanisms of action associated with Rezu-lin,” 91 but there is no indication that Dr. Bonkovsky intends to testify that Rezu-lin can cause liver injury silently.

6. Summary

To the extent the experts would testify that Rezulin can cause “silent” liver injury, they have postulated a causal chain in which Rezulin can affect the mitochondria or the BSEP and thus trigger apoptosis, which would be injurious but silent. A graphical representation of the theory might be useful:

[[Image here]]

*411 B. The Science upon Which the Experts Rely

The Court now will review the scientific research that the experts cite in support of their opinions. As will be discussed in a later section, the Court does not necessarily agree that all of the cited studies support the conclusions. The Court’s concern for the time being, however, is to present the experts’ own case. Nonetheless, at certain points in the following discussion, the Court will add its own observations— for example, comments on aspects of the studies, on the nature of the connections between the studies and the expert opinions, or on the absence of certain kinds of evidence — when it believes that additional information is essential to evaluate the testimony in question.

As an initial matter, the plaintiffs’ experts cite no studies that say that Rezulin can cause a silent liver -injury. There are no clinical trials and no observational epidemiological studies supporting the plaintiffs’ position. Nor have the plaintiffs pointed to any clinical case in which Rezu-lin was believed to have caused such an injury. 92 Rather, the opinions at issue on this motion are based on studies that the experts say support individual links in their causal chain. In particular, the experts have cited studies that connect Rezu-lin to apoptosis, studies that connect Rezu-lin to mitochondrial damage, and studies that connect Rezulin to an effect on the BSEP. Nearly all of these studies were conducted in vitro.

*412 1. Early Links in the Proposed Causal Chain: The Claim that Rezulin Causes Apoptosis Through Effects on the Mitochondria or the BSEP

a. Studies Connecting Rezulin to Apop-tosis

The Court begins with studies cited by the experts for the proposition that Rezu-lin caused apoptosis in certain kinds of cells.

At the outset, it is important to note that none of these studies dealt with normal human liver cells. The plaintiffs and their experts imply that Kostrubsky et al. (2000), 93 a study that found that Rezulin was toxic to normal human liver cells at certain concentrations, shows that Rezulin produces apoptosis. That study, however, says nothing at all about whether the cell death occurred via apoptosis or necrosis. 94 The same is true of several other studies cited by the plaintiffs and their experts, including Ramachandran et al. (1999). 95

The experts cite one study in which Rezulin was found to cause apoptosis in rat hepatocytes. Toyoda et al. (2001) found that Rezulin produced this effect at concentrations of 15 p.M 96 and above. 97 Furthermore, a Warner-Lambert scientist tentatively reached the same conclusion from the company’s own data. 98

The experts cite four 99 studies in which Rezulin was found to produce apoptosis in *413 human liver cancer cells. Bae and Song (2003), Toyoda et al. (2002), Yoshizawa et al. (2002), and Yamamoto et al. (2001) observed that Rezulin produced this result in a dose-dependent fashion above certain concentrations.

In addition, the experts cite a number of studies in which Rezulin at certain concentrations produced apoptosis in cultured cancer cells of other organs. Harris and Phipps (2002) found that troglitazone induces apoptosis in malignant white blood cells. Ohta et al. (2001) observed a similar result in thyroid cancer cells. Tsubouchi et al. (2000) found the same in lung cancer cells. Guan et al. (1999) did so in bladder cancer cells, and Takahashi et al. (1999) and Sato et al. (2000) obtained this result in stomach cancer cells. Finally, Dr. Reed cites a trio of cases on prostate cancer cells, but they did not find that Rezulin produces apoptosis in human cells. 100

Several of the studies on cancer cells examined the effect of Rezulin in concert with other substances. For example, the Kim et al. (2002) study examined the effect on lines of ovarian cancer cells of combining an anti-tumor agent known as TRAIL with troglitazone. The researchers found that troglitazone combined with TRAIL (but neither alone) killed tumor cells; the mechanism was believed to be apoptosis. Similarly, Elstner et al. (1998) found that the combination of troglitazone and another chemical (but neither by itself) caused apoptosis in human breast cancer cells.

Finally, the experts cite studies on what Dr. Reed describes as “normal” cells from other organs. 101 Gouni-Berthold et al. (2001) and Okura et al. (2000) obtained results suggesting that troglitazone caused apoptosis in cultures of a type of blood vessel cell taken from ráts and associated with atherosclerosis in humans. YamazaH et al. (2002) and Kawahito et al. (2000) derived a similar result for cultures of a type of cell that destroys the joints of patients with rheumatoid arthritis because it proliferates like a tumor. And Rovin et al. (2002) found that troglitazone induces cell death believed to be apoptotic in a type of kidney cell that proliferates in response to injury but that can cause kidney failure if the proliferation is excessive.

b. Studies Connecting Rezulin to Mitochondrial Damage

In addition to the studies that show apoptosis from Rezulin, albeit not in normal human liver cells, the experts cite studies showing that Rezulin can affect the mitochondria of liver cells. A preliminary clarification is called for. Toxicity from Rezulin. linked to mitochondrial damage is not necessarily relevant to the silent injury theory. It is uncontested that an injury would not be silent if it occurred by necrosis. The studies on mitochondria are relevant to the silent injury theory set forth in *414 the expert reports only to the extent that those studies can be used to connect Rezu-lin with apoptosis.

The experts cite a number of studies in which Rezulin caused morphologic, or structural, changes to mitochondria. Caldwell et al. (2001a) administered normal doses of Rezulin to NASH patients for six months and examined, among other things, their mitochondria. These investigators found that more of the mitochondria had a misshapen and unusually elongated appearance after treatment. They found as well “a decrease in the mean total number of mitochondria ... and an increase in the mean percentage of mitochondria that were enlarged and contained intramito-chondrial crystals ... but neither change was statistically significant.” 102

Although Drs. Smith’s and Reed’s expert reports do not mention it, these researchers looked for evidence of apoptosis. They examined the post-treatment biopsies of five patients with an electron microscope and found that “features of apoptosis (bleb formation, nuclear fragmentation) were not evident.” 103 They then applied a biochemical marker that detects apoptosis and found “[n]o statistically significant difference between the before and after specimens[.]” 104

Apart from Caldwell et al. (2001a), the other mitochondria studies all were performed in vitro. Shishido et al. (2003) applied troglitazone at concentrations up to 50 p,M — the researchers chose that concentration because it had prevented the cells from proliferating without inducing apoptosis in a previous study performed on liver cancer cells 105 — to an immortalized 106 cell line and found “marked enlargement” 107 and other abnormalities in the mitochondria. Tirmenstein et al. (2002) applied the drug to liver cancer cells and found “extensive alterations in mitochondrial morphology.” 108

Several of the studies cited by the experts measured a quantity known as the mitochondrial membrane potential, which is the electrical gradient, or polarization, between the outside and inside of a mito-chondrion. A decrease in this polarization indicates that the mitochondria are not functioning normally. Haskins et al. (2001) found that troglitazone at certain concentrations produced a decrease in the mitochondrial membrane potential in rat hepatocytes and in the hepatocytes of one diabetic human donor, but not in the cells of another diabetic donor. These researchers wrote that “[f]uture studies should address the downstream effects of these changes, since nuclear condensation and apoptosis are observed as late effects that are shared by” troglitazone’s chemical family. 109 Shishido et al. (2003) and Tir-menstein et al. (2002) found that troglita-zone at certain concentrations depolarized the mitochondrial membrane in the'immortalized liver cells and in liver cancer cells, respectively.

*415 Dr. Smith’s reports, but not Dr. Reed’s, argue that the above-described effects on the mitochondria link Rezulin to apoptosis. Dr. Smith begins with the statement that “[t]he decline in mitochondrial transmem-brane potential seen in these experiments is called the mitochondrial permeability transition.” 110 Dr. Smith continues:

“During the mitochondrial permeability transition, pores in the mitochondrial membrane are opened and [a protein known as] cytochrome c is released into the cytoplasm.... The cell ‘knows’ the cytochrome e should be normally be [sic ] inside the mitochondria, so when it léaks out it acts as a signal telling the cell that the mitochondria are badly damaged, and this stimulates the cell to commit suicide (i.e. apoptosis)[.]” 111

Dr. Smith here cites Ravagnan et al. (2002). That article 112 states that once a cell has been induced to die, “the outer mitochondrial membrane becomes completely permeabilized to proteins, resulting in the leakage of potentially toxic mitochondrial intermembrane proteins” — the authors include cytochrome c in this category — “that orchestrate the degradation phase of apoptosis.” 113 Ravagnan et al. (2002) thus does not say, as Dr. Smith seems to, that mitochondrial change (due to Rezulin or otherwise) alone can be an independent cause" of apoptosis. Rather, the authors view certain mitochondrial changes as themselves products of some prior cause of cell death and as one step in a chain of events that leads to apoptosis. Another article cited by Dr. Smith, Bissell et al. (2001), reflects a similar understanding. 114

A cited study that does agree with Dr. Smith 115 is Kim et al. (2003). These researchers found that the change in permeability of the mitochondrial membrane triggered cell death in rat hepatocytes. This paper, however, reports that the cell will undergo apoptosis — as opposed.to necrosis — only if there is sufficient ATP. The organized death of the cell is a process that consumes energy, and without energy *416 (in the form of ATP), the cell defaults to the necrotic mode of death. 116

Another paper that agrees that the depolarization of the mitochondrial membrane itself can be a cause of cell death is Tirmenstein et al. (2002). This study, however, does not indicate whether the cell death occurs via apoptosis or necrosis, and it specifies that troglitazone leads to a decrease in the amount of ATP in the cell. Furthermore, the concentrations at which troglitazone produced cell death in this study were higher than those at which it began to affect the mitochondria.

Dr. Smith continues:

“A protein in the cytoplasm, called Bax, then migrates to the mitochondrial membranes and accelerates the apoptotic process by making the mitochondrial membrane even more permeable, thus further releasing even more cytochrome c into the cytoplasm thereby speeding up the apoptotic process[.]” 117

Here Dr. Smith cites Smaili et al. (2001). Those researchers found that a decline in the mitochondrial membrane potential precedes the attachment of Bax to the mitochondria in cells genetically engineered to overproduce Bax. The authors sum up the implications of their research this way: “our results suggest that alterations in mitochondrial energization associated with apoptosis can initiate Bax docking to mitochondria.” 118 The study, in other words, says nothing about whether mitochondrial dysfunction (whether caused by Rezulin or otherwise) is an independent cause of apoptosis. This study, like the research reviewed in Ravagnan et al. (2002), created conditions known to induce apoptosis and then examined the sequence of events within the cell.

Dr. Smith concludes:

“For TRO induced liver cell apoptosis, this process has recently been studied in detail by Bae and Song.... They found a critical role for Bax and [a protein known as JNK] activation in TRO induced liver cell apoptosis.... TRO ... induced apoptosis that was preceded by activation of JNK ... and increased levels of Bax [and] release of cytochrome c....” 119

Bae and Song (2003), as mentioned above, found that troglitazone causes apoptosis in liver cancer cells. They found as well that TRO increases the levels of proteins associated with apoptosis, including Bax and cytochrome c. They did not, however, examine the mitochondria.

c. Studies Connecting Rezulin to an Effect on the BSEP

Dr. Smith’s other proposed mechanism through which Rezulin allegedly could cause silent injury involves the BSEP. As with the evidence on mitochondria, and for the same reasons, any effect on the BSEP is relevant to the silent injury theory only to the extent that it connects Rezulin with apoptosis.

The experts cite several articles in which Rezulin was found to interfere with the ability of cells to export bile salts. Preminger et al. (1999) reported that troglita-zone interfered with the flow of bile in isolated rat livers. Two studies by Funk et al. confirmed this result in live rats; they found that when troglitazone was administered to rats, there was an increase in the concentration of bile acid in the rats’ *417 blood plasma believed to have been caused by interference with the BSEP. Elevation of bile acid is analogous to elevation of bilirubin; 120 both indicate a cholestatic injury.

To connect the presumed effect on the BSEP to apoptosis, Dr. Smith and Dr. Reed cite a group of studies and review articles. 121 All are concerned with elucidating the biochemical mechanisms by which the build-up of bile in the liver can produce apoptosis.

2. The Last Link in the Chain: The Claim that Apoptosis from Rezulin Causes Silent Injury

The plaintiffs’ experts have cited no evidence that any apoptosis caused by Rezu-lin can result in a silent injury. That is, the plaintiffs’ experts have offered no evidence that apoptosis caused by Rezulin, if any, occurs at a clinically significant level or, even if it does, that the resulting injury would remain silent.

The following exchange occurred during Dr. Smith’s testimony:

“The Court: I think ... it is your view ... that doses of Rezulin that have been shown to be cytotoxic in cell studies obviously can be achieved in some human cells. We agree so far?

“The Witness: Sure. I agree on it.

“The Court: ... [W]hat, if anything, does that tell us on an empirical basis, as opposed to a theological basis, [about] whether achievement of those doses in some undefinable proportion of liver cells has any serious adverse consequences for anybody? ... Isn’t it entirely possible, and indeed probable, that there is some minimal number or proportion of liver cells that have to be killed off before there is any interference with the fiver’s functioning that matters to anybody?

“The Witness: That is true.

“The Court: Right. Do we know what this, and how it relates to the effect of Rezulin?

“The Witness: I don’t think we do know, your Honor.” 122

Dr. Reed testified in a similar way. The following exchange occurred during his cross-examination:

“Q. Doctor, you testified that you believe it is possible that Rezulin could be inducing a low level of apoptosis in human fivers, is that correct?

“A. Yes, I believe that is possible.

“Q.... [D]o you have any study or any data to support that sustained apoptosis has resulted in clinical injury in a patient taking Rezulin?

“A. Such a study has not been performed, to my knowledge.” 123

Dr. Reed further testified as follows:

“The Court: So just to try to put it in bottom line colloquial terms, even assuming ... everything you and Dr. Smith have said about the capability of Rezulin to induce apoptosis in particular fiver cells at what I will call for the sake of quickness relevant concentrations, we have no empirical data that indicates to us how many or what percentage of the fiver cells are in *418 jured in that way in vivo and what the threshold level of injury, in terms of numbers or proportion of liver cells, needs to be before there is any clinical injury to the patient?

“The Witness: No, we really don’t know the percentage. All we know is, as I have said, one out of every 50 people has sufficient cell death occurring ... that their transaminase levels rose when they took the drug.” 124

C. Arguments that Rezulin Can Cause Silent Liver Injury Through Mechanisms Other than Apoptosis

In addition to the apoptosis theory, the plaintiffs’ counsel have argued that liver injury caused by Rezulin can be silent either because Rezulin suppresses the production of enzymes in the liver or because the enzymes might have been depleted already in a patient with pre-existing liver injury. 125 The logical implication is that even if the apoptosis theory discussed above were unsubstantiated or incorrect, plausible mechanisms by which Rezulin can cause silent liver injury would remain. The problem is that these mechanisms, for the most part, were proposed by the plaintiffs’ counsel, not their experts. To the limited extent that they have been adopted by the experts, they are not supported. 126 The only physiological explanation offered by the plaintiffs’ experts in support of *419 their opinion that Rezulin can cause liver injury silently that requires further consideration is that involving apoptosis.

D. Patients Whose Liver Enzymes Were Not Monitored

Plaintiffs’ counsel raise the possibility that some people had abnormally elevated liver enzymes while on Rezulin therapy that were not recorded, either because the patient was not tested or was not tested at a time when the enzymes were elevated. 127 The Court recognizes this possibility. But it has no bearing on the present motion, as it does not speak to whether Rezulin can cause liver injury through a mechanism that does not result also in elevated enzymes.

V. Law Governing the Admission of Expert Testimony

A. Daubert and Its Progeny

The standard governing a district court’s determination whether to admit scientific or other expert testimony is familiar. Federal Rule of Evidence 702 provides:

“If scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness qualified as an expert by knowledge, skill, experience, training, or education, may testify thereto in the form of an opinion or otherwise, if (1) the testimony is based upon sufficient facts or data, (2) the testimony is the product of reliable principles and methods, and (8) the witness has applied the principles and methods reliably to the facts of the case.”

This rule incorporates principles established by the Supreme Court in Daubert v. Merrell Dow Pharmaceuticals, Inc. 128 The Court there recognized that district judges have a “gatekeeping” 129 role in which they must ensure that “scientific testimony or evidence admitted is not only relevant, but reliable.” 130 Daubert observed that the relaxation for expert witnesses of the usual requirement of first-hand knowledge presumably “is premised on an assumption that the expert’s opinion will have a reliable basis in the knowledge and experience *420 of his discipline.” 131 For that reason, a trial judge must “make certain that an expert ... employs in the courtroom the same level of intellectual rigor that characterizes the practice of an expert in the relevant field.” 132

A federal district judge faced with a challenge to the admissibility of expert testimony:

“must determine at the outset ... whether the expert is proposing to testify to ... scientific knowledge that ... will assist the trier of fact to understand or determine a fact in issue. This entails a preliminary assessment of whether the reasoning or methodology underlying the testimony is scientifically valid and of whether that reasoning or methodology properly can be applied to the facts in issue.” 133

These matters are to be determined by a preponderance of proof. 134

The inquiry is “a flexible one.” 135 In Daubert, the Supreme Court articulated four pertinent factors while leaving open the possibility of others: (1) whether the expert’s theory “can be (and has been) tested;” (2) whether the theory “has been subjected to peer review and publication;” (3) the “known or potential rate of error;” and (4) whether the theory has “widespread acceptance.” 136 Courts have considered other factors as well, including whether an expert’s opinion was developed for litigation and whether the expert has accounted adequately for obvious alternative explanations. 137

In addition, Rule 702’s requirement that the proposed testimony “assist the trier of fact to ... determine a fact in issue” means that the proffered testimony must be “sufficiently tied to the facts of the case that it will aid the jury in resolving a factual dispute.” 138 The Daubert decision recognized that this consideration, which has been described as one of “fit,” “goes primarily to relevance.” 139 A district court therefore is not required “to admit opinion evidence that is connected to existing data only by the ipse dixit of the expert. A court may conclude that there is simply too great an analytical gap between the data and the opinion proffered.” 140

*421 Thus in General Electric Co. v. Joiner, 141 the Supreme Court held that a district court did not abuse its discretion when it excluded the testimony of experts that exposure to polychlorinated biphenyls (PCBs) was likely responsible for the lung cancer of an electrician who was a smoker and had a family history of lung cancer. The Supreme Court found that the animal studies on which the experts relied, which involved exposing infant mice to massive doses of PCBs, “were so dissimilar to the facts presented in this litigation that it was not an abuse of discretion for the District Court to have rejected the experts’ reliance on them.” 142 The Supreme Court likewise did not take issue with the district court’s conclusions that four epidemiological studies were not a reliable basis for the experts’ opinions. In two of the four studies, the authors were unwilling to conclude that PCB exposure had caused cancer in the observed workers. The third study did not even involve PCBs, and the subjects in the fourth had been exposed to numerous carcinogens in addition to PCBs. 143

The Second Circuit’s recent case law follows the logic of Joiner and Daubert. In Amorgianos v. National Railroad Passenger Corp., 144 the Circuit reiterated the familiar principles governing the admissibility of expert testimony and further held that:

“[t]o warrant admissibility ... it is critical that an expert’s analysis be reliable at every step.... In deciding whether a step in an expert’s analysis is unreliable, the district court should undertake a rigorous examination of the facts on which the expert relies, the method by which the expert draws an opinion from those facts, and how the expert applies the facts and methods to the case at hand.” 145

In that case the district court was confronted with expert testimony that, among other things, exposure to certain toxic chemicals was a general cause of the type of neuropathy allegedly suffered by the plaintiff bridge painter. The district court “conducted an extremely thorough review of the scientific literature” on which those experts relied. 146 The Circuit found that this review:

“was certainly within the broad discretion afforded to the district court under Daubert and its progeny, and did not impinge upon the jury’s function. It is precisely such an undertaking that assures that an expert, when formulating an opinion for use in the courtroom, will employ the same level of intellectual rig- or as would be expected in the scientific community.” 147

The Circuit upheld the district court’s exclusion of the experts. 148 The district court found that the articles cited by at least one of the experts “fail[ed] to ‘fit’ the facts of this case, either in terms of the type and duration of exposure, or the type and duration of the observed effects.” 149 It found it significant as well that this *422 expert “prepared his opinion for litigation, rather than as part of his academic research, and he has not seen fit to share his opinion” with the relevant professional community. 150

The Second Circuit’s most recent pronouncement on the admissibility of expert testimony came last year in Wills v. Amerada Hess Corp, 151 In that case, the Circuit held that a district court did not abuse its discretion when it excluded an expert report theorizing that a sailor’s cancer probably was caused by toxins to which he had been exposed on the defendant’s ships. The district court found that the studies on which the expert had relied did not establish a sufficient link between the toxins and the type of cancer because they were done only on animals and found a significant link only when the animals ingested the chemical, which the sailor had not done. The court was dissatisfied also with the expert’s failure to account adequately for the possibility that the cancer had been caused by the sailor’s smoking and drinking. 152 Finally, the court found that the expert’s novel theory of causation — which the expert admitted was “the product of his own ‘background experience and reading’ ” 153 — failed the Daubert test because it had not been tested or subjected to peer review, there was no known error rate, and it was not generally accepted. 154

B. The Daubert Standards Apply to Opinions About General and Specific Causation

As discussed earlier, a plaintiff in a toxic tort case must prove both general causation, that is, that the alleged toxin is capable of causing injuries of the kind suffered by the plaintiff, and specific causation, that is, that the alleged toxin caused the particular plaintiffs injuries. The Daubert requirements apply alike to expert opinions on general and specific causation. Indeed, since Daubert many district courts‘have excluded expert testimony regarding general causation on grounds quite similar to those proposed here. 155

*423 VI. Daubert Analysis of the Proposed Testimony

A Testing and Error Rate, Peer-Review, Publication, Widespread Acceptance

The challenged testimony in this case does not satisfy any of the core Daubert factors. The theory that Rezulin can cause a liver injury silently never has been tested 156 and necessarily has no error rate. It never has been published or subjected to peer review — aside from an edited version of Dr. Smith’s report, which used more tentative language 157 and which he published in a toxicology journal at the suggestion of a member of its editorial board who also is a paid consultant for the plaintiffs in this litigation. 158 It appears to have no acceptance outside this litigation, let alone widespread acceptance. 159

*424 The plaintiffs argue that' the research linking Rezulin to apoptosis and apoptosis to injury is part of the established scientific literature. 160 The challenged testimony however, is the opinion that Rezulin is capable of causing a silent liver injury. At issue, in other words, is much more than the independent assertions that Rezulin causes apoptosis in some types of cells under some conditions and that excessive apoptosis can be injurious. It is the extrapolation from the existing literature that never has been tested, peer-reviewed, published, or widely accepted. 161 As the Second Circuit has made clear, “it is critical that an expert’s analysis be reliable at every step ... ‘any step that renders the analysis unreliable under the Daubert factors renders the expert’s testimony inadmissible.’ ” 162

B. Independence from Litigation

There is likewise no real dispute that the theory that Rezulin can cause a silent liver injury was developed solely in connection with this litigation. The plaintiffs have come forward with no evidence that Dr. Smith ever proposed this idea publicly before, and Dr. Reed admitted at the evi-dentiary hearing that he had not presented the theories in his expert report on the possible mechanisms of Rezulin’s alleged toxicity in any context other than this litigation. 163

*425 C. Consideration of Contrary Evidence

A factor that courts have considered in Daubert analyses is whether an expert has accounted adequately for obvious alternative explanations. This is appropriate because any theory that fails to explain information that otherwise would tend to cast doubt on that theory is inherently suspect. By the same token, if the relevant scientific literature contains evidence tending to refute the expert’s theory and the expert does not acknowledge or account for that evidence, the expert’s opinion is unreliable. Accordingly, courts have excluded expert testimony “where the expert selectively chose his support from the scientific landscape.” 164

In this case, the plaintiffs’ experts have ignored a large amount of information that calls many aspects of the silent injury theory into question.

Their reports do not mention that the one study that looked for Rezulin-induced apoptosis in humans failed to find it. Caldwell et al. (2001a) administered Rezu-lin to human patients and looked for morphological and biochemical evidence of apoptosis but found none. 165

The expert reports point out that Toyoda et al. (2001) found that Rezulin induced apoptosis in healthy rat hepatocytes. But they do not mention that Kim et al. (2002) — a paper of which Dr. Reed was a coauthor and which he discussed in his declarations — investigated healthy hepato-cytes from cynomologus monkeys and found that Rezulin, alone or in combination with the anti-tumor agent TRAIL, failed to induce1 apoptosis. Rats are much further than monkeys from humans, and Dr. Reed himself regards cells from monkey as a useful alternative to human cells. 166

The expert reports do not mention that four of the studies on cancer cells — including the Kim et al. (2002) study performed in Dr. Reed’s laboratory — compared the effect of troglitazone on the cancer cells to the effect on healthy cells and that, in every such comparison, troglitazone was found to induce apoptosis in the cancer cultures while having no effect on the healthy cells. 167 The expert reports do not mention, in other words, that the reason *426 many researchers were investigating tro-glitazone’s effects on cancer cells and on diseased tissue is that they were actively exploring the possibility that the drug could be an effective therapy for killing or preventing the proliferation of malignant tissue without harming healthy cells.

The expert reports do not discuss the fact that the same researchers who found that troglitazone affected the mitochondria in the cells of live humans, immortalized liver cells, and liver cancer .cells either observed no evidence at all of apoptosis or other cell death 168 or, if they did observe cell death, (i) observed it at concentrations higher than the concentrations required to affect the mitochondria, and (ii) did not specify whether the mechanism was apop-tosis. 169 The expert reports did not discuss the fact that Shishido et al. (2003), a study performed on immortalized human liver cells, looked for but could not find cytochrome c, considered a sign of apopto-sis, and believed this absence “suggest[ed] no clear involvement of mitochondria-mediated apoptosis in this in vitro system as previously reported.” 170 The expert reports saw fit to omit that, while Haskins et al. (2001) did find that troglitazone affected the mitochondria in liver cells derived from one diabetic patient, these researchers failed to see this result in the cells derived from another diabetic patient.

In other words, the scientists have discussed only the evidence that they believed would advance the plaintiffs’ position. Their reports cannot be said to reflect “the same level of intellectual rigor that characterizes the practice of an expert in the relevant field.” 171

D. “Fit” and the “Analytical Gap”

A crucial consideration in evaluating the admissibility of expert testimony is whether the conclusions flow reliably from the premises. As the Supreme Court has explained, “[a] court may conclude that there is simply too great an analytical gap between the data and the opinion proffered.” 172 Or as Judge Becker has explained for the Third Circuit: “[e]ven if an expert’s proposed testimony constitutes scientific knowledge, his or her testimony will be excluded if it is not scientific knowledge for purposes of the case.” 173 This consideration is bound up with the relevancy requirement described in Daubert as one of “fit.”

The analytical gap between the research and the conclusions the experts would draw is independently sufficient to warrant exclusion of the testimony in question. The plaintiffs have no evidence for the final link in their causal chain, and they extrapolate from the earlier links in ways the Court finds unreliable.

1. The Experts Have No Evidence for the Crucial Link in Their Causal Chain

The experts have no evidence to carry them all the way down their causal chain to silent liver injury.

*427 It is worth bearing in mind that some level of apoptosis is entirely normal and occurs all the time in healthy tissue. Drs. Reed and Smith admitted that they have no information on whether therapeutic doses of Rezulin, even assuming they can cause additional levels of apoptosis beyond the normal baseline, can do so to a clinically relevant extent.

Furthermore, both sides’ experts and the literature agree that when an insult producing apoptosis is sufficiently serious, the injury or impairment will not be silent. Thus the question is not just whether Re-zulin can cause apoptosis to occur at levels sufficient to cause injury or impairment, but whether it can do so and remain silent. The plaintiffs’ experts have offered no evidence that this is possible. Indeed, nothing in the challenged expert reports gives any indication about the actual capacity of phagocytes (the “clean-up” cells) to absorb the end products of apoptosis — -the precondition to the injury’s hypothesized silence.

11. The Experts Have Failed To Link the Studies on Mitochondria and the BSEP into Their Causal Chain

Similarly, the studies on mitochondria have no connection to Rezulin-induced apoptosis. Dr. Smith would opine that (a) Rezulin has been shown to affect the structure and function of the mitochondria, and (b) apoptosis has been shown to involve changes to the mitochondria, and therefore (c) Rezulin can produce apoptosis. This is speculative, and it confuses association with causation. Dr. Smith’s reasoning is logically equivalent to saying that (a) every time John gets hungry he eats, and (b) John eats .whenever he goes to a restaurant, therefore (c) every time John gets hungry, he goes to a restaurant. There are of course plenty of times when John gets hungry and eats at home.

So too here. There is not simply the possibility that Rezulin changes the mitochondria without producing apoptosis. There are empirical demonstrations of this in the studies of Caldwell et al. (2001a), 174 Shishido et al. (2003), 175 and Tirmenstein et al. (2002) 176 performed, respectively, on live humans, immortalized human liver cells, and liver cancer cells. 177 The plain *428 tiffs’ experts must have some reliable basis for asserting that Rezulin-induced mitochondrial abnormalities lead specifically to apoptosis, otherwise the mitochondria research does nothing to help them survive this motion. The plaintiffs’ experts have no such basis.

Nor is Dr. Smith’s theory rendered reliable by the fact that he, and some of the articles he cites, assert that a change in mitochondrial membrane function itself is an initial cause of, and not just a phenomenon associated with, apoptosis. The fact that there is compelling evidence (which he ignores) in the very studies that he and Dr. Reed cite to the effect that human liver cells — normal, immortalized, and cancerous' — can sustain mitochondrial damage without shriveling up and dying undermines what the plaintiffs’ counsel have described as Dr. Smith’s “one plus one equals two” 178 argument.

The BSEP research suffers from the same basic difficulty discussed above. That is, while there are studies to suggest that Rezulin, through a mechanism involving the BSEP, can produce cholestatic injury and that cholestatic injury is associated with apoptosis, there is no reason to believe that any apoptosis caused by BSEP malfunction would be silent. The cholestatic injury that Rezulin produced in rats in the two Funk et al. studies was manifested by the presence in the rats’ blood of products closely associated with bilirubin.

For apoptosis associated with cholestasis to be silent, surrounding phagocytes would have to absorb the dying bile salt-filled cells. Can they really do this? The very toxicity to cells of bile and related compounds — this is to be contrasted with ALT and AST — is a starting point of the research on the BSEP and cholestasis. Dr. Smith offers no help in resolving this paradox, and therefore no reason to believe that inhibition of the BSEP by Rezulin could produce a silent injury.

3. The Research on Apoptosis in Cell Cultures Does Not “Fit” the Opinion at Isstoe

The preceding discussion shows that the most that can be said for the research cited by the experts, to the extent it is relevant at all, is that it shows that at certain concentrations, Rezulin causes apoptosis in the hepatoeytes of rats (but not monkeys), in human liver cancer cells, in cancerous cells of other organs (but, so far as the data presented to the Court indicate, not in healthy cells from those organs), and in non-cancerous disease tissue. 179 The plaintiffs argue that the experts can extrapolate reliably from in vitro results to draw conclusions about the effect of Rezulin in humans.

Caution always must be used in extrapolating results in tissue culture to effects in *429 live humans. 180 As explained in the Reference Manual on Scientific Evidence:

“Thousands of in vitro toxicological tests have been described in the scientific literature .... There are short-term in vitro tests for just about every physiological response and every organ system .... Relatively few of these tests have been validated by replication in many different laboratories or by comparison with outcomes in animal studies to determine if they are predictive of whole-animal or human toxicity.” 181

In assessing the reliability of an extrapolation from in vitro results to effects in live humans, two crucial considerations are the type of cell on which the in vitro experiment was performed and the dose to which the cells were exposed. 182

a. Types of Cells

In the cited studies, the cells in which Rezulin was found to produce apoptosis were not normal human liver cells. They were either healthy liver cells of rats, cancerous human liver cells, and cancerous or otherwise abnormal cells from other human organs. This fact is of substantial significance. In light of the explanations of the plaintiffs’ experts and the concessions of the plaintiffs’ counsel, almost none of these studies appears to be a rehable basis for extrapolating to the liver of a living human.

To begin with, Dr. Smith described at the hearing a “hierarchy” of relevance of experiments on different kinds of cells. 183 At the top of the hierarchy is human hepa-toeytes. Next is “immortalized” hepato-cytes, which Dr. Smith rates “about equivalent” to “rat hepatocytes or any of the animal liver cells.” 184 The third level is “liver tumor cells, such as hepatoma cell lines. And the fourth part of the hierarchy would be non-target tissue cells.” 185 The suggestion is that extrapolation becomes less and less appropriate as one proceeds down the hierarchy.

In fact, it is on the basis of this testimony that the plaintiffs’ counsel have argued that Rezulin’s pattern of harming malignant but not normal tissue from various organs does not detract from their hypothesis that Rezulin is toxic to the (non-cancerous) liver. In their proposed findings of fact, the defendants pointed out that in the body of science before the Court, every *430 time the effects of Rezulin on cancer cells were compared with the effects on normal cells, Rezulin killed the cancer cells without damaging the normal ones. 186 The plaintiffs’ response, for each of these studies, is some form of the statement that “[t]he experiment did not concern liver cells and is therefore of limited relevance to troglitazone hepatotoxicity.” 187 Indeed, the plaintiffs make a similar statement in connection with other studies performed on non-liver cells. 188

The plaintiffs thus effectively have abandoned all of the studies on non-liver cells as a basis for their experts’ opinions. Accordingly, the Court is convinced that most of the in vitro studies upon which the plaintiffs’ experts initially relied do not support a reliable extrapolation to the opinion that Rezulin can cause a silent liver injury in humans.

Furthermore, the Court does not regard the four cited studies in which troglitazone produced apoptosis in human liver cancer cells as a reliable basis for predicting the drug’s effect on non-cancerous human livers. The plaintiffs and their experts have ignored the evidence that Rezulin does not have an apoptotic effect on non-cancerous monkey liver cells and failed to deal with the fact that cancer cells — in the four studies cited to this Court that compared them with healthy cells from the same organ— respond to troglitazone in ways that healthy cells do not. And it is undisputed that cancer cells are abnormal precisely in that their mechanisms of cell growth and death have been disturbed. 189 Indeed, Dr. Smith himself ranked cancerous liver cells as third out of fourth in his hierarchy of relevance.

When all of the studies involving human liver tumor cells and cells from other organs are stripped away, what is left is the Toyoda et al. (2001) study, which investigated rat hepatocytes and is considered below.

b. Dose

It is a fundamental principle of toxicology, as Dr. Smith explained at the hearing, that “the dose makes the poison.” 190 Consequently, if the doses at which Rezulin was observed to be toxic to cultured cells are not achieved in the liver in vivo, extrapolation from the in vitro experiments *431 is not reliable. 191 In light of the preceding discussion, the only in vitro study that the Court finds it necessary to consider further is Toyoda et al. (2001), which investigated rat hepatocytes. 192 Thus the question is whether the doses to which the cells in that study were exposed are comparable to those to which cells in the liver of a living human are exposed.

Dose is a function both of concentration of a toxin and of time of exposure. 193 Toyoda et al. (2001) found that troglitazone at concentrations of 15 pM and above killed most of the rat hepatocytes within 20 hours. But that result obtained only in the absence of albumin, a protein to which Rezuhn is more than 99 percent bound in human blood. 194 The binding to albumin means that the molecules of troglitazone are not as free to interact with other substances. Results obtained in the absence of albumin therefore cannot readily be generalized or extrapolated to a living human. 195

Toyoda et al. (2001) itself demonstrates the significance of the presence of albumin. When those researchers added a 2 percent solution of bovine serum albumin to the cells, the same 15 pM concentration of troglitazone that was otherwise so toxic to the cells produced virtually no effect after 24 hours. Even a 100 pM solution of troglitazone destroyed less than 20 percent of the cells, and a 50 pM solution destroyed approximately 5 percent. 196 In human blood plasma, the concentration of albumin is 4 percent, 197 or double the maximum used in the Toyoda et al. (2001) study. The question raised by the Toyoda et al. (2001) results therefore is whether the concentrations at which troglitazone produced apoptosis in the presence of albumin in vitro are comparable to the concentrations of troglitazone achieved in vivo in therapeutic settings.

When a patient was taking prescribed doses of Rezulin, the average maximum concentration in the blood plasma was determined to be 2.0 to 6.3 pM, depending on the dosage, 198 a finding substantially adopted by experts on both sides. 199 The defendants argue that the concentration in the liver of a patient taking Rezulin is approximately equal to that in the blood plasma, 200 which is a small fraction of the *432 concentrations of troglitazone found to be toxic in the Toyoda et al. (2001) study.

The plaintiffs’ experts contend that the concentration of Rezulin inside a patient’s liver actually is much higher — 50 to 80 pM, according to Dr. Smith. 201 At the evidentiary hearing, Dr. Smith said the disparity between this level and the concentration found in the bloodstream is attributable to hepatic first-pass uptake, which is the liver’s absorption of certain substances from the bloodstream right after they have "left the gut. 202 Furthermore, the plaintiffs’ experts have cited research that they say corroborates this range. 203

Before examining Dr. Smith’s analysis, the Court pauses to point out several major difficulties with extrapolation from the published literature. The concentration of troglitazone discussed in Toyoda et al. (2001) is the concentration in the medium that surrounds the cells, not the concentration inside the liver cells. However, Kawai et al. (1997), the lone published study that found that the concentration of troglitazone in the liver is higher than in the plasma, 204 sought to measure the concentration in the entire liver tissue without distinguishing between the concentrations inside and outside the cells. 205

These two concentrations are not interchangeable. The plasma membrane is a selective barrier. Indeed, it is the cell’s ability to maintain internal conditions that differ from those outside that makes all life possible. 206 As one of the defense- experts has explained:

“Intact cells have a complex membrane separating the intracellular contents from the outside medium or blood. The degree to which a substance in the extracellular medium traverses the cell membrane is a complex function of many factors.... It is thus grossly incorrect to adopt an intracellular concentration, even if correct, and apply that to phenomena observed at a specific extracellular concentration.” 207

*433 Furthermore, it is unknown whether any toxicity due to troglitazone is a function of its presence outside the plasma membrane of the cell (where it may interact with certain proteins embedded in the membrane) or inside the cell (where it can interact with compounds in the cytoplasm).

Another difficulty is that the concentrations measured in Kawai et al. (1997) actually encompassed not just the pure trogli-tazone molecule, but the various related molecules, known as metabolites, that result from biochemical breakdown processes inside the cell. 208 Dr. Smith could give little information the extent to which these metabolites are toxic. 209 In consequence, the concentrations observed by Kawai et al. (1997) may overstate the relevant concentrations by aggregating troglitazone and metabolites that may have different properties.

Thus, although it may be true that tro-glitazone and its metabolites accumulate in the liver, there is still no basis for the assertion that it is clinically realistic to expose cells in culture to extracellular concentrations comparable to the whole-tissue concentrations alleged by Drs. Smith and Reed. The question, contrary to what the plaintiffs and their experts have implied, is not whether the concentrations used in Toyoda et al. (2001) match the combined concentration of troglitazone and its metabolites across the entire liver; rather, it is whether the concentrations of troglitazone used in Toyoda et al. (2001) match the concentrations of troglitazone to which the outside of the cells in patients actually were exposed. Even putting this point aside, however, there are independent difficulties in Dr. Smith’s projection of the concentration of Rezulin in the liver based on absorption on the “first pass.”

Dr. Smith started with the premise that Rezulin’s absolute “bioavailability” — the fraction of the ingested drug that is absorbed from the gut instead of remaining there 210 — is 50 to 70 percent. In other *434 words, if a patient takes 600 milligrams (mg) of Rezulin, 300 to 420 mg will be absorbed by the body. 211 The Court notes that the 50 percent figure is comparable to, if on the high end of, published estimates. 212 The 70 percent figure is based on the premise that food increases bioa-vailability. 213

Dr. Smith next stated that 43 percent of that amount (i.e. 129 to 180.6 mg) will be absorbed directly by the liver on the first pass from the gut. 214 He cited Izumi et al. (1996) for that figure, but the study does not support it. In fact, Izumi et al. (1996) assumed what the plaintiffs dispute; which is that “the. concentrations of unbound drug in venous blood' and liver are equal.” 215 Nevertheless, using the molecular weight of troglitazone (441 grams/ mole) and assumptions about the mass of the liver (1500 grams) and its density (1 gram/1.08 milliliters), Dr. Smith converted the 129 to 180.6 mg to a concentration range: 180.5 to 252.8 |xM. 216

But this number is meaningless. 217 Dr. Smith himself explained that such a concentration would obtain only if all of the Rezulin absorbed by the liver on the first pass were absorbed at once, which it is not, rather than over time. 218 Thus, Dr. Smith said, “If you do a toxicokinetics model of troglitazone it predicts the concentration of troglitazone in the liver will be 50 to 80 micromolar after a therapeutic concentration....” 219

The plaintiffs have failed to establish any reliable basis for the 50 to 80 p,M range. Dr. Smith gave no information on how he reached that number. While he methodically explained the calculations and assumptions that produced the range 180.5 to 252.8 jjlM, he simply uttered “toxicoki-netic modeling” before producing the range of 50 to 80 p,M. He provided no information concerning the structure of the model or the data and assumptions used in producing the claimed result.

Moreover, even if Dr. Smith’s assertion on this point were accepted, the Court could not accept the figure as reliable. For one thing, it is based on the premise that 43 percent of the absorbed fraction of troglitazone is absorbed directly by the liver on the “first pass,” a number that the Court could not find in the only study that Dr. Smith cited for it. And there is still the problem that Dr. Smith failed to explain how the tissue-wide concentrations he postulates can be used to relate (a) the extracellular concentrations to which cells actually were exposed in patients taking Rezulin to (b) the concentrations used in Toyoda et al. (2001).

In spite of all of this, even if it were accepted that the concentration of troglita-zone in the liver is 50 to 80 p,M, and even if the failure to distinguish between intracellular and extracellular concentration were overlooked, there would be another reason why extrapolation from the Toyoda et al. (2001) study cannot be said to be *435 reliable. The experts have provided no information on the distribution of troglita-zone in the liver. Drs. Smith, Reed, and Julie agree that Rezulin is not distributed homogeneously in the liver, 220 but they cannot say how it is distributed. 221 That is, they have no information on the percentage of cells that would be exposed to the concentrations that they assert are toxic. The drug may concentrate in discrete pockets, perhaps close to the portal vein, 222 causing localized damage but leaving the rest of the organ unharmed and consequently causing no clinically significant effect. The plaintiffs’ experts have no answer to this difficulty.

The Court finds that extrapolation from the results of Toyoda et al. (2001) to the livers of patients who took Rezulin is unwarranted. The Court cannot and does not say that Drs. Smith and Reed are wrong or that the defendants’ experts are right. Rather, it is not satisfied that there is a reliable basis for an extrapolation from the results of an in vitro experiment that used troglitazone at a concentration of 50 or 100 p,M with a 2 percent solution of albumin to clinical effects in a patient taking Rezulin. Similar statements could be made about other studies relied on by the experts, but the Court finds it unnecessary to discuss the specifics of those studies. 223

VII. The Plaintiffs’ Argument Concerning Differential Diagnosis

The plaintiffs attempt to get around all of these problems by arguing that the testimony in question could be used to support opinion testimony regarding causation based on the clinical process of elimination known as differential. diagnosis. 224

The Federal Judicial Center’s Reference Manual on Scientific Evidence has explained differential diagnosis this way:

*436 “[T]he physician determines which of two or more diseases with similar clinical findings is the one that the patient is suffering from. The physician does this by developing a list of all of the possible diseases that could produce the observed signs and symptoms, and then comparing the expected clinical findings for each with those exhibited by the patient.” 225

The plaintiffs argue that the opinions about silent liver injury are admissible because they are based on what they call a “plausible mechanism.” 226 The plaintiffs’ position is that a physician faced with a patient who took Rezulin and experienced silent liver injury could attribute that injury to Rezulin by performing a differential diagnosis that takes that “plausible mechanism” into account. 227 But the plaintiffs have underestimated their burden under Daubert.

A physician attempting to establish a causal relationship between exposure to a substance and a particular patient’s illness must “demonstrate that the medical and scientific literature provides evidence that in some circumstances the exposure under consideration can cause the outcome under consideration. This step is synonymous with establishment of general causation.” 228 In other words, the determination of cause in an individual case, or differential diagnosis as that term has been used by courts, “does not ‘speak to the issue of general causation. [It] assumes that general causation has been proven for the list of possible causes’ that it rules in and out in coming to a conclusion.” 229 As one court has explained:

“the final, suspected ‘cause’ remaining after this process of elimination must actually be capable of causing the injury.... And, of course, expert opinion on this issue of ‘general causation’ must be derived from scientifically valid methodology.” 230

A physician thus may not link any particular patient’s injury to Rezulin unless there is some reliable basis for the opinion that therapeutic doses of Rezulin can cause such an injury.

The Court is mindful that two district judges in this Circuit seem to have come to a different view, stating, with reference to the Second Circuit’s decision in McCullock *437 v. H.B. Fuller Co., 231 that “[differential diagnosis is a reliable basis to prove general causation in this circuit” 232 and that “according to the Second Circuit, if a qualified expert performs a reliable differential diagnosis, the plaintiff need not satisfy the general causation requirement.” 233 But the view that differential diagnosis necessarily is sufficient to establish general causation is not borne out by McCulloch. The Circuit there merely registered its approval of the expert’s reliance on a variety of sources to arrive at an opinion as to causation in one.patient’s case, an opinion that, so far as the Circuit’s opinion indicated, did not differentiate between general and specific causation. It is not at all clear that the Court regarded differential diagnosis as inevitably probative of general causation. 234

The plaintiffs point out that researchers and clinicians in the field of liver toxicology use a variety of data and analysis' — -sometimes including differential diagnoses on individual patients — to conclude that a particular drug is capable of causing a particular injury. 235 The plaintiffs further point out, correctly, that the physicians who first determined that Rezulin caused certain types of liver injuries (non-silent ones) had no published studies but made judgments based on a totality of information. 236 The implication is that physicians retained for this litigation should be able to testify that Rezulin is a possible, and the most likely, cause of particular plaintiffs’ symptoms— symptoms that were consistent with liver injury but that occurred without elevated enzymes — and that these physicians should be able to do so by relying upon, among other things, the expert testimony that the defendants are seeking to exclude on this motion.

The plaintiffs’ mistake throughout is to overlook the fact that science and medicine refine themselves over time,’ gradually converting tentative hypotheses into reliable principles. The idea that Rezulin sometimes causes acute liver injury, albeit not silently, appears not now to be controversial for a number of reasons, including the existence of a significant number of compelling case reports and several generally accepted, empirically verified (though not necessarily perfectly understood) physiological processes through which drug-induced liver injury occurs. A medical theory that has no little or no empirical support, is entirely unaccepted outside the very lawsuit in which the theory is being advanced for the first time, and that suffers from numerous analytical gaps is an entirely different matter.

VIII. Conclusion

To sum up, the plaintiffs have not established the reliability of the silent injury theory. The theory never has been tested or peer-reviewed, has not been published except by Dr. Smith after the commencement of this litigation and. only then in speculative terms and suspicious circumstances, and has no acceptance outside this litigation. The plaintiffs’ experts have ignored information that appears to call crucial aspects of their theory into question. The theory rests on a series of empirically unbridgeable analytical gaps. Most impor *438 tantly, the experts have not established a sound basis for concluding that Rezulin-induced apoptosis can occur at clinically significant levels and remain silent. Similarly, the experts have failed to show that any mitochondrial changes attributable to Rezulin themselves can cause apoptosis or that any cholestatic injury due to Rezulin’s effect on the BSEP would be silent.

Thus, all the experts really have are a set of studies in which Rezulin produced apoptosis in tissue cultures. These studies, however, do not “fit” the opinion they are used to support. Save one study on rat hepatocytes, all were performed on cancerous liver cells and cancerous or otherwise unhealthy cells of other organs. The plaintiffs’ counsel and their experts have acknowledged that studies on such cells are not reliable predictors of in vivo outcomes in a non-cancerous liver.

Nor is the study on rat hepatocytes a reliable basis for extrapolation. Among other issues, there is no reason to believe that the doses used in that study (or, for that matter, in many of the other studies presented to the Court) approximated the doses to which clinically relevant quantities of cells in the human liver are exposed.

The plaintiffs attempt to deal with all of these problems by arguing that the testimony in question could factor into the diagnosis of an individual patient. The plaintiffs’ position is that a physician, faced with a patient who took Rezulin and had symptoms of liver disease but no elevated enzymes, could use the opinions of Drs. Smith and Reed and the research they cite to conclude that the patient’s injury was caused by Rezulin. The flaw, however, is that a physician must have some reliable basis for believing that a particular substance is capable of causing the injury in question in relevant circumstances before concluding that the substance caused that injury in a particular case. Here, there is no such basis.

:¡í :¡: $ ‡ ‡ $

In a now-famous passage, the Supreme Court recognized that the trial judge’s ga-tekeeping role:

“inevitably on occasion will prevent the jury from learning of authentic insights and innovations. That, nevertheless, is the balance that is struck by Rules of Evidence designed not for the exhaustive search for cosmic understanding but for the particularized resolution of legal disputes.” 237

The Court is fully aware that investigation of the molecular mechanisms of cell death is one of the most promising and potentially clinically relevant frontiers in science. Indeed, time and medical science may one day prove the plaintiffs’ experts right. But in Judge Posner’s oft-quoted words, “the courtroom is not the place for scientific guesswork, even of the inspired sort.” 238

Federal Rule of Evidence 702 and Dau-bert and its progeny set forth standards and give district courts the obligation and the discretion to determine whether expert testimony meets those standards. The Court is convinced that the silent injury theory does not. The defendants’ motion to exclude proposed testimony of plaintiffs’ experts that Rezulin can cause liver injury, or exacerbate a pre-existing liver condition, in the absence of marked elevations of liver enzymes while the patient was taking the medication [docket item 899] therefore is granted.

SO ORDERED.

*439 Appendix: Scientific and Medical Studies Discussed in the Opinion

Myung-Ae Bae & Byoung J. Song, Critical Role of c-Jun N-Terminal Protein Kinase Activation in Troglitazone-Induced Apoptosis of Human HepG2 Hepatoma Cells, 63 Molecular Pharmacol. 401 (2003) (Ex. 10). Bae & Song (2003)

D. Montgomery Bissell et al., Drug-Induced Liver Injury: Mechanisms and Test Systems, 33 Hepatol. 1009 (2001) (Ex. 17). Bissell et al. (2001)

Rachel Butler et al., Nonapoptotic Cell Death Associated with S-Phase Arrest of Prostate Cancer Cells via the Peroxisome Proliferator-Activated Receptor K Ligand, 15-Deoxy-K-12’u-prostaglandin J2 11 Cell Growth & Differentiation 49 (2000) (Ex. 26). Butler et al. (2000)

Stephen H. Caldwell et al., A Pilot Study of a Thiazolidinedione, Troglitazone, in Nonalcoholic Steatohepatitis, 96 Am. J. Gastroenterol. 519 (2001) (Ex. 27). Caldwell et al. (2001a)

Stephen H. Caldwell, Letter to the Editor, Efficacy and Safety of Troglitazone for Lipodystrophy Syndromes, 134 Annals Internal Med. 1008 (2001) (Ex. 31). Caldwell (2001b)

Elena Elstner et al., Ligands for Peroxisome Proliferator-Activated Receptor y and Retinoic Acid Receptor Inhibit Growth and Induce Apoptosis of Human Breast Cancer Cells In Vitro and in BNX Mice, 95 Proc. Nat’l Acad. Sci. 8806 (1998) (Ex. 49). Elstner et al. (1998)

William A. Faubion et al., Toxic Bile Salts Induce Rodent Hepatocyte Apoptosis via Direct Activation of Fas, 103 J. Clinical Investig. 137 (1999) (Ex. 58). Faubion et al. (1999)

Miya Fukano et al., Subacute Hepatic Failure Associated with a New Antidia-betic Agent, Troglitazone: A Case Report with Autopsy Examination, 31 Human Pathol. 250 (2000) (Ex. 65). Fukano et al. (2000)

Christoph Funk et al., Cholestatic Potential of Troglitazone as a Possible Factor Contributing to Troglitazone-Induced Hepatotoxicity: In Vivo and In Vitro Interaction at the Canalicular Bile Salt Export Pump (Bsep) in the Rat, 59 Molecular Pharmacol. 627 (2001) (Ex. 67). Funk et al. (2001a)

Christoph Funk et al., Troglitazone-Induced Intrahepatic Cholestasis by an Interference with the Hepatobiliary Export of Bile Acids in Male and Female Rats, 167 Toxicol. 83 (2001) (Ex. 68). Funk et al. (2001b)

Ioanna Gouni-Berthold et al., Troglitazone and Rosiglitazone Induce Apoptosis of Vascular Smooth Muscle Cells Through an Extracellular Signal-Regulated Kinase-Independent Pathway, 363 Naunyn-Schmiedeberg’s Arch. Pharma-col 215 (2001) (Ex. 78). Gouni-Berthold et al. (2001)

You-Fei Guan et al., Expression of Peroxisome Proliferator-Activated Receptor (PPARy) in Human Transitional Bladder Cancer and Its Role in Inducing Cell Death, 1 Neoplasia 330 (1999) (Ex. 84). Guan et al. (1999)

Sarah G. Harris & Richard P. Phipps, Prostaglandin D2 Its Metabolite 15-d-PGJ2 and Peroxisome Proliferator Activated Receptor-y Agonists Induce Apoptosis in Transformed, but Not Normal, Human T Lineage Cells, 105 Immunol. 23 (2002) (Ex. 85). Harris & Phipps (2002)

Jeffrey R. Haskins et al., Thiazolidinedione Toxicity to Isolated Hepatocytes Revealed by Coherent Multiprobe Fluorescence Microscopy and Correlated with Multiparameter Flow Cytometry of Peripheral Leukocytes, 75 Arch. Toxicol 425 (2001) (Ex. 87). Haskins et al. (2001)

Steven K. Herrine & Cuckoo Choudhary, Letter to the Editor, Severe Hepato-toxicity Associated with Troglitazone, 130 Annals Internal Med. 163 (1999) (Ex. 91). Herrine & Choudhary (1999)

*440 Takashi Izumi et al., Prediction of the Human Pharmacokinetics of Troglita-zone, a New and Extensively Metabolized Antidiabetic Agent, after Oral Administration, with an Animal Scale-up Approach, 277 J. Pharmacol. & Exp. Therapeutics 1630 (1996) (Ex. 98). Izumi et al. (1996)

Hartmut Jaeschke et al., Forum: Mechanisms of Hepatotoxicity, 66 Toxicol. Sci. 166 (2002) (Ex. 100). Jaeschke et al. (2002)

Yutaka Kawahito et al., 15-deoxy~A.12’li-PGJ2 Induces Synoviocyte Apoptosis and Suppresses Adjuvant-Induced Arthritis in Rats, 106 J. Clinical Inves-tig. 189 (2000) (Ex. 108). Kawahito et al. (2000)

Kenji Kawai et al., Disposition and Metabolism of the New Oral Antidiabetic Drug Troglitazone in Rats, Mice and Dogs, 47 Arzneim.-Forsch7Drug Res. 356 (1997) (Ex. 109). Kawai et al. (1997)

Jae-Sung Kim et al., Mitochondrial Permeability Transition in the Switch fi'om Necrotic to Apoptotic Cell Death in Ischemic Rat Hepatocytes, 124 Gastroenterol. 494 (2003) (Ex. 114). Kim et al. (2003)

Youngsoo Kim, Nanjoo Suh, Michael Sporn & John C. Reed, An Inducible Pathway for Degradation of FLIP Protein Sensitizes Tumor Cells to TRAIL-Induced Apoptosis, 277 J. Biol. Chemistry 22320 (2002) (Ex. 115). Kim et al. (2002)

Hironori Koga et al., Involvement ofp21WAFIICiv\ p27Kip', and p!8iNK4c in Troglitazone-Induced Cell-Cycle Arrest in Human Hepatoma Cell Lines, 33 Hepatol. 1087 (2001) (Ex. 117). Koga et al. (2001)

James Kohlroser et al., Hepatotoxicity Due to Troglitazone: Report of Two Cases and Review of Adverse Events Reported to the United States Food and Drug Administration, 95 Am. J. Gastroenterol. 272 (2000) (Ex. 118). Kohlroser et al. (2000)

Vsevolod E. Kostrubsky et al., The Role of Conjugation in Hepatotoxicity of Troglitazone in Human and Porcine Hepatocyte Cultures, 28 Drug Metabolism & Disposition 1192 (2000) (Ex. 119). Kostrubsky et al. (2000)

Tetsuya Kubota et al., Ligand for Peroxisome Proliferator-Activated Receptor K (Troglitazone) Has Potent Antitumor Effect Against Human Prostate Cancer Both In Vitro and In Vivo, 58 Cancer Res. 3344 (1998) (Ex. 120). Kubota et al. (1998)

Paul Kwo et al, Nuclear Serine Protease Activity Contributes to Bile Add-Induced Apoptosis In Hepatocytes, 268 Am. J. Physiol. G613 (1995) (Ex. 123). Kwo et al. (1995)

John J. Lemasters, Mechanisms of Hepatic Toxicity V. Necrapoptosis and the Mitochondrial Permeability Transition: Shared Pathways to Necrosis and Apoptosis, 276 Am. J. Physiol.-Gastrointestinal & Liver Physiol. G1 (1999) (Ex. 128). Lemasters (1999)

John J. Lemasters et al., The Mitochondrial Permeability Transition in Cell Death: A Common Mechanism In Necrosis, Apoptosis and Autophagy, 1366 Biochimica et Biophysica Acta 177 (1998) (Ex. 127). Lemasters et al. (1998)

Cho-Ming Loi et al., Steady-State Pharmacokinetics and Dose Proportionality of Troglitazone and Its Metabolites, 39 J. Clinical Pharmacol. 920 (1999) (Ex. 137). Loi et al. (1999a)

Cho-Ming Loi et al., Clinical Pharmacokinetics of Troglitazone, 37 Clinical Pharmacokinetics 91 (1999) (Ex. 136). Loi et al. (1999b)

Hideyuki Miyoshi et al., Hepatocyte Apoptosis After Bile Duct Ligation in the Mouse Involves Fas, 117 Gastroenterol. 669 (1999) (Ex. 154). Miyoshi et al. (1999)

*441 Elisabetta Mueller et al., Effects of Ligand Activation of Peroxisome Prolifera- Mueller et al. (2000) tor-Activated Receptor y in Human Prostate Cancer, 97 Proc. Nat’l Acad. Sci. 10990 (2000) (Ex. 156).

Elizabeth J. Murphy et al., Troglitazone-Induced Fulminant Hepatic Failure, Murphy et al. (2000) 45 Digestive Diseases & Sci. 549 (2000) (Ex. 158).

Kazuyasu Ohta et al., Ligands for Peroxisome Proliferator-Activated Receptor Ohta et al. (2001) K Inhibit Growth and Induce Apoptosis of Human Papillary Thyroid Carcinoma Cells, 86 J. Clinical Endocrinol. & Metabolism 2170 (2001) (Ex. 162).

Takafumi Okura et al., Troglitazone Induces Apoptosis via the p5S and Gaddis Okura et al. (2000) Pathway in Vascular Smooth Muscle Cells, 407 Eur. J. Pharmacol. 227 (2000) (Ex. 163).

Tushar Patel et al., Apoptosis and the Liver: A Mechanism of Disease, Growth Patel et al. (1999) Regulation, and Carcinogenesis, 30 Hepatol. 811 (1999) (Ex. 166).

Kurt Preininger et al., Acute Troglitazone Action in Isolated Perfused Rat Preminger et al. Liver, 126 British J. Pharmacol. 372 (1999) (Ex. 175). (1999)

Vinod Ramachandran et al., Troglitazone Increases Cytochrome P-150 SA Ramachandran et al. Protein and Activity in Primary Cultures of Human Hepatocytes, 27 Drug (1999) Metabolism & Disposition 1194 (1999) (Ex. 178).

Luigi Ravagnan et al., Mitochondria, the Killer Organelles and Their Weapons, Ravagnan et al. (2002) 192 J. Cellular Physiol. 131 (2002) (Ex. 181).

Robert Routh et al, Troglitazone Inhibits Glutamine Metabolism in Rat Routh et al. (2002) Mesangial Cells, 282 Am. J. Physiol.-Endocrinol. & Metabolism E231 (2002) (Ex. 185).

Brad H. Rovin et al., 15-Deoxy-A12’li-^prostaglandin J2 Regulates Mesangial Rovin et al. (2002) Cell Proliferation and Death, 61 Kidney Int’l 1293 (2002) (Ex. 186).

J. Sahi et al., Effect of Troglitazone on Cytochrome PU50 Enzymes in Primary Sahi et al. (2000) Cultures of Human and Rat Hepatocytes, 30 Xenobiotica 273 (2000) (Ex. 187).

H. Sato et al., Expression of Peroxisome Proliferator-Activated Receptor Sato et al. (2000) (PPAR)y in Gastric Cancer and Inhibitory Effects of PPARy Agonists, 83 British J. Cancer 1394 (2000) (Ex. 188).

Shoichiro Shishido et al., Hydrogen Peroxide Overproduction in Megamito- Shishido et al. (2003) chondria of Troglitazone-Treated Human Hepatocytes, 37 Hepatol. 136 (2003) (Ex. 198).

S.S. Smaili et al., Bax Translocation to Mitochondria Subsequent to a Rapid Smaili et al. (2001) Loss of Mitochondrial Membrane Potential, 8 Cell Death & Differentiation 909 (2001) (Ex. 201).

Thomas Sodeman et al., Bile Salts Mediate Hepatocyte Apoptosis by Increasing Sodeman et al. (2000) Cell Surface Trafficking of Fas, 278 Am. J. Physiol.-Gastrointestinal & Liver Physiol. G992 (2000) (Ex. 206).

Nobuhiko Takahashi et al., Activation of PPARy Inhibits Cell Growth and ’ Takahashi et al. (1999) Induces Apoptosis in Human Gastric Cancer Cells, 455 FEBS Letters 135 (1999) (Ex. 211).

Mark Tirmenstein et al., Letter to the Editor (Reply to Vsevolod E. Kostrubsky Tirmenstein et al. et al., Letter to the Editor (critique of Tirmenstein et al. (2002))), 71 Toxicol. (2003) Sci. 282 (2003) (Ex. 219).

*442 Mark A. Tirmenstein et al, Effects of Troglitazone on HepG2 Viability and Tirmenstein et al. Mitochondrial Function, 69 Toxicol. Sci. 131 (2002) (Ex. 218). (2002)

M. Toyoda et al., A Ligand for Peroxisome Proliferates Activated Receptor 7 Toyoda et al. (2002) Inhibits Cell Growth and Induces Apoptosis in Human Liver Cancer Cells, 50 Gut 563 (2002) (Ex. 220).

Y. Toyoda et al., Toxic Effect of Troglitazone on Cultured Rat Hepatocytes, 68 Toyoda et al. (2001) Life Sci. 1867 (2001) (Ex. 221).

Yasunori Tsubouchi et al., Inhibition of Human Lung Cancer Cell Growth by Tsubouchi et al. (2000) the Peroxisome Proliferator-Activated Receptor-y Agonists Through Induction of Apoptosis, 270 Biociiem. and Biophys. Res Communications 400 (2000) (Ex. 224).

Yui Yamamoto et al., Cytotoxicity and Apoptosis Produced by Troglitazone in Yamamoto et al. Human Hepatoma Cells, 70 Life Sci. 471 (2001) (Ex. 239). (2001)

Ryuta YamazaH et al., Nonsteroidal Anti-Inflammatory Drugs Induce Apopto- Yamazald et al. (2002) sis in Association with Activation of Peroxisome Proliferator-Activated Recepten* 7 in Rheumatoid Synovial Cells, 302 J. Pharmacol & Exp. Therapeutics 18 (2002) (Ex. 241).

Kaname Yoshizawa et al., Peroxisome Proliferator-Activated Receptor 7 Li- Yoshizawa et al. (2002) gand Troglitazone Induces Cell Cycle Airest and Apoptosis of Hepatocellular Carcinoma Cell Lines, 95 Cancer 2243 (2002) (Ex. 243).