dissenting.
The- fundamental issue in this case is whether there is sufficient scientific and medical evidence for a jury to determine that Bryan Glaser’s brain injury, which was discovered after, and certainly resulted at least in part from, a fall, was proximately caused by his ingestion of a standard 75-milligram dose of Dexatrim, an over-the-counter diet pill. For the reasons given below, the evidence in support of that proposition is almost wholly speculative, and cannot justify submission of the issue to a jury.
*979I
This case represents a somewhat typical example of a broad category of tort litigation. The plaintiff engages in an allegedly causative act. In this case it is the taking of a non-prescription diet pill, with an active ingredient that Americans consume in the same dosage billions of times every year. The resultant event, which undoubtedly occurred, is a fall leading to the striking of plaintiffs head on the floor of the bank in which he was standing. Thus, we have an alleged cause, which is experienced annually by tens of millions of people. We have an alleged effect, a fall, which is also suffered by many, many people. The difficult question is what connects these two events, other than post hoc propter hoc. Also, as in most tort cases of this type, when there is little or no obvious cause and effect relationship, there is an expert witness on whom the plaintiff relies to present the case to the jury.
A
In this case, Dr. Zaloga is generally qualified as an expert in endocrinology, and has carried out, or been involved in, or reviewed scientific literature concerning, studies on the relationship between phenylpropanolamine (PPA), the active ingredient in Dexatrim; elevated blood pressure; and intracranial bleeding. The problem, however, is that Dr. Zaloga wishes to extrapolate from areas in which studies have shown some causative tendency into areas where the scientific data have in fact been negative or, at most, purely speculative. For example, some correlation, implying causation to some, has been shown when subjects take 150 mg of PPA, which is twice the dose that plaintiff is posited to have taken. Alternatively, some effect has been shown when subjects take 75 mg of PPA in connection with 400 mg of caffeine, an amount of caffeine equal to several cups of coffee or many cola drinks.1 Despite loose language in Zaloga’s deposition, which has been faithfully mined by the court’s opinion, the record contains neither well-documented clinical cases nor experimental studies showing a connection between 75 mg of PPA taken alone, and blood pressure high enough to result in intracranial bleeding and a stroke.
Second, it is important to note that the alternative and obvious explanation, that Bryan Glaser fainted and subsequently injured his head during the fall, is amply supported by the record. He had been dieting aggressively, had not been feeling well, had exhibited flu-like symptoms, had been examined and diagnosed with “post-viral syndrome” the very morning of his fall, and had been standing in a bank line for up to 25 minutes while wearing a coat inside a crowded, hot, and stuffy room just prior to his falling. Dr. Sharon, who later visited Glaser at the hospital at Bryan’s family’s request, explained the logic behind the assumption that Glaser suffered from “a viral syndrome and simply passed out ... because of prolonged standing in a line motionless while sick”:
A number of reasons could produce that: inflammation of his semi-circular canals causing vertigo, dehydration from the illness itself, whether it’s intestinal or just heavy breathing causing fluid loss from the lungs. Just no one — many times no one knows why you’re dizzy when the virus is in you.
J.A. at 442.
It is also important to note that Dr. Zaloga never examined the plaintiff, neither before nor after he fell; that the plaintiff was examined by Dr. Baubie of MedStop shortly before he fell, and his blood pressure was found to be within the normal range; and that Dr. Bodzin, the emergency room physician at Detroit Macomb Hospital who first examined Glaser subsequent to the fall, attributed the bleeding in his brain to the trauma of the fall, not to a pre-existing intracranial bleed. J.A. at 199.
*980B
Under Daubert v. Merrell Dow Pharmaceuticals, Inc., — U.S. -, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), and our court’s opinion in Turpin v. Merrell Dow Pharmaceuticals, Inc., 959 F.2d 1349 (6th Cir.), cert. denied, — U.S. -, 113 S.Ct. 84, 121 L.Ed.2d 47 (1992), a trial court is not required to allow a jury to pass on every case where a person with adequate general credentials can be found to utter some version of the talismanic phrase: “In my opinion it is more likely than not (or 80% probable) that X was caused by Y.” Rather, such testimony must be supported by a reasonable chain of evidence, not speculation, and it is subject to the scrutiny of the trial court. Daubert, — U.S. at -, 113 S.Ct. at 2798-99 (confirming that trial judges must play a “gatekeeping role” and should grant summary judgment when the “scintilla of evidence presented supporting a position is insufficient to allow a reasonable juror to conclude that the position more likely than not is true”); Turpin, 959 F.2d at 1352, 1359-60 (endorsing the “hard look doctrine”).
“[Ajlthough judges should respect scientific opinion and recognize their own limited scientific knowledge, nevertheless courts have a duty to inspect the reasoning of qualified scientific experts to determine whether a ease should go to the jury.” Turpin, 959 F.2d at 1350. As a general principle, courts look to medical experts to offer insights into medical “probabilities,” not mere “possibilities.” Tabaczynski v. United States, 529 F.Supp. 156, 163 (E.D.Mich.1981), aff'd mem., 711 F.2d 1059 (6th Cir.1983).
The plaintiff in a pharmaceutical product-liability action needs to establish that the drug under attack “more probably than not caused” the injury. Turpin, 959 F.2d at 1359. Thus, if the probabilities are “at best evenly balanced,” Michigan law still directs a verdict for the defendant. Mulholland v. DEC Int’l Corp., 432 Mich. 395, 416 n. 18, 443 N.W.2d 340, 350 n. 18 (1989) (citation omitted). It does not suffice to produce evidence that the drug “is ‘capable of causing,’ ‘could cause’ or its effects are ‘consistent with causing’ ” the injury. Turpin, 959 F.2d at 1360. Rather, the plaintiff must show that “it probably causes [that kind of injury] in general or that it did in this case.” Ibid. Thus, a verdict for a plaintiff who claimed that she had contracted rheumatoid arthritis from swine-flu vaccine was overturned by this court because the plaintiff had only shown that “an antibody antigen reaction can cause rheumatoid arthritis, [but there was] no showing that it did cause the plaintiffs disease in this case.” Hasler v. United States, 718 F.2d 202, 205 (6th Cir.1983) (emphasis added), cert. denied, 469 U.S. 817, 105 S.Ct. 84, 83 L.Ed.2d 31 (1984). This threshold is not met when a medical expert testifies to mere personal belief or opinion.
We believe that close judicial analysis of such technical and specialized matter is necessary not only because of the likelihood of juror misunderstanding, but also because expert witnesses are not necessarily always unbiased scientists. They are paid by one side for their testimony.[2] Although there is no suggestion of unethical scientific conduct in the present case, the potential for exaggeration and fraud on the court is present and may be impossible to discover without close inspection and careful consideration of the record. As Judge Leventhal observed in the context of administrative law, in some circumstances there exists a “combination of danger signals” requiring enhanced “judicial vigilance to enforce the Rule of Law.” In such situations, “a court does not depart from its proper function when it undertakes a study of the record, hopefully perceptive, even as to evidence on technical and specialized matters....”
Id. at 1352-53 (citations omitted).
C
In this ease, under the rubric of “differential diagnosis,” Dr. Zaloga gave learned disquisitions as to why it was implausible that certain arcane conditions, such as major nervous systemic damage, genetic defects, or etiopathic bleeding, might possibly have caused the fall. He ruled out the possibility that Glaser’s fall was caused by untreated *981hypertension, cocaine use, a tumor, an aneurism, leukemia, hemophilia, and Von Willer-brand’s disease. However, with regard to the obvious possibility that Glaser simply fainted after standing in a stagnant bank line for as long as twenty-five minutes while wearing a coat inside a crowded, hot, and stuffy room, a short time after going to a doctor complaining of flu-like conditions, and during a period when he was dieting aggressively and eating improperly, Dr. Zaloga lightly brushed aside the plausibility of such an explanation with the extremely conelusory statement that “I don’t have any evidence presented to me that suggests that he slipped or he fell. He wasn’t in line for a prolonged period of time to have le[]d to pooling. We have no evidence that he was severely dehydrated or other reason to have fallen.” J.A. at 365. Another portion of his testimony is further revealing:
THE WITNESS: I believe the teller testified — again, this is to my recollection — that she saw him sweating and thought it was unusual enough — different from anyone else — and did grab his head and looked somewhat out of it.
Q. (Mr. Lichtman) Do you know how hot it was in the bank?
A. No.
Q. Do you know whether or not it was warmer than it otherwise might have been in the bank?
A. I’m presuming that it — this is January 4th, 1988, and that it wasn’t that hot. I mean, I’m — I’m just guessing on the time of the year. I’m also guessing that •people — I believe there — there are descriptions of coats. I can’t recollect that that’s true or not.
And the teller described him as looking much different than the other people. So I — I don’t know the exact answer to the question, but it would be hard for me to believe that it was so overly hot that he would be the only one to experience symptoms and no one else would be — at least by the teller’s description.
But I don’t know the specific answer to how hot it was in the bank.
Depo.Tr. at 128-29 (emphasis added). Thus, Dr. Zaloga discounted a teller’s eyewitness account and personal recollection of conditions in the bank, all contained in a vivid description that he had read within the month.3 Later, when asked again concerning the facts pertinent to this case, Dr. Zaloga responded:
Q. Now, if you consider that a person is wearing an overcoat in an inside of a bank that is warm and is waiting about 20, 25 minutes, standing up, on line, has already seen a doctor for fatigue that day, has told the doctor that they have not been feeling well, the doctor diagnoses the person as having a post-viral syndrome, and there is evidence that the person has not been eating a lot and — or potentially fasting that day, do you believe that that could lead to someone fainting?
THE WITNESS: My — I think it’s possible, but very unlikely. We have people like that all the time....
Depo.Tr. at 182-83.
Instead, without explaining how he arrived at such numbers, he simply assigned a “probability” of 80% to PPA ingestion as the cause of the fall and 15% to the fall being caused by fainting.4 J.A. at 402-03. Probability is a legal term of art. It should be based on scientific method, not on personal opinion. See, e.g., Turpin, 959 F.2d at 1359-60. Nowhere does Zaloga’s testimony explain where the numbers “80” and “15” come from. On the contrary, the numbers become fuzzier with the passage of a day’s testimony:
So my opinion is that, in my mind, I would put the most likely cause for this injury to have been an intercerebral bleed which caused the fall, rather than some other event. I’m still leaving there, you *982know — I said before — maybe an 80 percent probability of that sequence.
And I’m leaving two other diagnoses, as I said yesterday, that it’s possible that he had the fall and then bled, and I — I have put a 15 percent on that, I believe, and a five percent for other etiopathie unknown causes.
Depo.Tr. at 300 (emphasis added). The court rejected such testimony in Turpin, where a medical witness testified with similarly unsupported imprecision, and this court stated that “[w]e cannot find ... that this testimony is anything more than a personal belief or opinion- Indeed, no understandable scientific basis is stated. Personal opinion, not science, is testifying here.” 959 F.2d at 1360.
D
Dr. Zaloga prefaced his assignment of percentages with comments based on his general belief that even 75 mg of PPA can cause clinical damage from hypertension.5 However, the cases in the medical literature to which Zaloga cites involve instances in which subjects were administered either 150 mg of PPA, or a combination of 75 mg of PPA with very substantial doses of caffeine. In the studies cited, no clinically significant effect was found, despite the very small number of subjects studied, in groups taking 75 mg alone.
The fallacy of the court’s conclusion may be underscored by the implication that follows. Every day, many Americans faint and fall. By Dr. Zaloga’s analysis, every one of them can get to a jury if they happen to have taken any of the 125-plus over-the-counter common-cold remedies or diet aids that provide doses of 75 mg of PPA (that is, not just Dexatrim, but also such medications as Con-tac and Dimetapp) and have no other strong organic reason for a faint.
The only caveat to this sweeping result would be Dr. Zaloga’s statements that (1) Glaser was a healthy person who did not have flu symptoms (despite Dr. Baubie’s diagnosis at MedStop, and Dr. Zaloga’s own positing later that, as an alternative explanation to his preferred one, perhaps Glaser “did have a viral type syndrome ... some type of fatigue syndrome that led him to be weak and he fell based on that,” Depo.Tr. at 206); and that (2) the pattern of brain damage did not seem to him to be consistent with damage from a fall (though it did seem so to Dr. Bodzin, the emergency room physician who first treated Glaser after the incident).6
However, Zaloga did not examine Glaser at any time, and the court does not indicate whether these two considerations are crucial, or whether the studies on PPA alone would have “carried the day.” Moreover, the court does not indicate whether it considers Zalo-ga’s type of long-distance diagnosis (clearly influenced heavily by the fact that Zaloga deeply believes that PPA can cause strokes in many consumers) to be an adequate basis for the crucial point that flu and other conditions did not cause the fall and the bleed.
In short, all that we have is a witness’s personal belief that an environmental condition to which tens of millions are exposed annually is capable of wreaking random havoc. This belief is not supported by the scientific studies he refers to. Nevertheless, he believes and is willing to testify that an adverse effect following the condition is caused by the condition, and he minimizes very obvious alternative explanations.
This is post hoc propter hoc reasoning at its rankest, and is contrary to the “hard look” encouraged and even required by our case law. See, e.g., Turpin, 959 F.2d at 1352. The trial judge in this case did exactly what he should have done in his “gatekeeping role” by taking such a hard look. His actions warrant praise, not reversal.
*983II
The court’s analysis of the evidence for its first proposition, that 75 mg of PPA is capable of causing hypertension severe enough to result in an intereranial bleed, can be divided into three parts: (1) the Zaloga/Lake studies; (2) the Horowitz and Pentel studies; and (3) the adverse case reports.
These materials must be considered against the background of the significant studies, on large samples, introduced by the defense. Dr. Blackburn conducted a study with over 800 subjects. George L. Blackburn, et al., “Determinants of the Pressor Effect of Phenylpropanolamine in Healthy Subjects,” 261 JAMA 3267 (1989), reprinted in J.A. at 310. Dr. Noble studied nearly 300 subjects. Rudolf Noble, “A Controlled Clinical Trial of the Cardiovascular and Psychological Effects of Phenylpropanolamine and Caffeine,” 22 Drug Intelligence & Clinical Pharmacy 296 (1988), reprinted in J.A. at 316. Dr. Liebson studied 150 subjects. Ira Liebson, “Phenylpropanolamine: Effects on Subjective and Cardiovascular Variables at Recommended Over-the-Counter Dose Levels,” 27 J. Clinical Pharmacology 685 (1987), reprinted in J.A. at 321. These studies showed statistically reliable but clinically insignificant increases in blood pressure with just 75 mg of PPA.
For example, Blackburn showed average blood pressure increases of 2-4 mm Hg with 75 mg of PPA, as opposed to a placebo. These findings were regarded as clinically insignificant. J.A. at 311, 313. Although he would like to see more individual data, Zalo-ga himself recognizes the Blackburn study as “a large-scale study, I think it’s a reasonably good study.” Depo.Tr. at 495. Some studies showed virtually no statistical effect at all from PPA. See Noble, J.A. at 316-18; Lieb-son, Table I, J.A. at 325. Zaloga recognizes both of these studies in terms similar to his description of the Blackburn study. Depo. Tr. at 502, 504. Despite the court s citation of Zaloga’s critique that these studies report their results only in terms of mean and do not provide variances or ranges, the studies cited in this paragraph report findings just as do the Zaloga/Lake studies, reporting mean, variance, and statistical significance. See, e.g., J.A. at 312, 317, 324-25.7
Ill
In contrast to these large studies, the “five papers” by Drs. Zaloga, Lake, and others are apparently based on only two studies, each involving strikingly small subject samples usually comprised of the same few participants. The first paper cited by the court, at 973 n. 2, involved only six participants. J.A. at 289. The third paper, at 973 n. 4, involved only sixteen subjects, six of whom were the same subjects who had comprised the first paper’s sample. Depo.Tr. at 47. Those same sixteen appear as the identical test sample in the fourth paper, at 974 n. 6, and once again in the fifth paper, at 974 n. 7. Compare Table I, J.A. at 295, with Table II, J.A. at 478, and J.A. at 464 & 473 n. 9.8 A careful look at the essence of the Zalo-ga/Lake reported data, in terms of both mean values and individual data points and peak ranges relied on by the authors, reveals that, ultimately, there is no evidence that the ingestion of a standard 75 mg dose of PPA alone produces other than clinically trivial results.
A
The first paper specifically found that 75 mg of PPA “did not cause clinically relevant hypertension.” J.A. at 289. This paper also defined hypertension as blood pressure of 140/90, and noted that some 16 billion doses of PPA are taken annually in the United States, and 20% of the United States population takes over 700 mg of caffeine daily.
*984B
The second paper again specifically found that during all the times that the subjects were tested, there were no clinically significant blood pressure rises, though the authors speculated that since they did not test at all times, something could have happened at a time they were not testing — an obvious tautology. The authors even candidly stated: “We conclude that 75 mg of sustained release PPA has only minimal effects ... in young, healthy, normotensive subjects.” J.A. at 461. The actual average values show this finding:
PLACEBO 75 MG PPA
SUPINE 116/72 117/75
STANDING 110/76 116/80
See Table I, J.A. at 460.
C
With the third paper, the story becomes a bit more interesting. The court correctly states that the paper reports that “single doses of PPA may significantly elevate blood pressure, usually at approximately 3 to 4 hours after ingestion.” At 973. However, an examination of the full paper reveals additional salient points.
First, the paper summarized its “Results,” reporting that “significant BP [blood pressure] increases occurred over several hours following 150 mg PPA and after 75 mg PPA plus 400 mg caffeine.” J.A. at 294. The word significant, as it is used in this context, means that the increases were “statistically significant” in that the increases, whatever their magnitude, probably did not happen by chance. The summary of “Results” further reported specifically that “[significant BP increases after ingestion of 75 mg PPA and after 400 mg caffeine were less frequent.” Ibid. The “CONCLUSION” section reported only that “150 mg PPA ... substantially elevates BP.” As used here, substantially indicates an amount of elevation. Consequently, the “Conclusion” ended with the authors’ recommendation that physicians inform patients “of the risks of taking more than the recommended amounts of PPA and of combining caffeine with PPA” Ibid, (emphasis added).
Second, even an examination of the detailed results shows an inconclusive picture. At J.A. 297, supported by Table II at J.A. 296, we find the statement that the average value for systolic blood pressure (SBP) was “significantly higher” (meaning statistical significance) at Hours 3 and 4; but diastolic blood pressure (DBP) was higher only at Hour 4. On the other hand, the peak value for SBP after 75 mg of PPA was not significantly different from that found after a placebo, but the peak value for DBP was significantly higher. No statement is made as to the clinical relevance of these differences. The magnitude of the change can be seen from Table II, as follows, at Hours 3 and 4, the times with the greatest difference:
PLACEBO 75 MG PPA 150 MG PPA
HOUR 3 124/75 136/85 157/94
HOUR 4 125/73 134/83 144/88
This table shows, even with the tiny number of subjects, which permits greater variability, that 75 mg PPA did not result in average values that were even minimally into the hypertensive range of 140/90, while 150 mg PPA did.
Finally, the court notes, at 973, that three subjects taking 75 mg of PPA “had [SBP] readings of 160mm Hg or higher.” However, this result is exactly the same as for those taking 400 mg of caffeine and no PPA. J.A. at 297.
D
The fourth paper again used the faithful 16 subjects. As reported by the court, there were increases in both SBP and DBP. However, as the authors themselves noted, “[A]ll peak pressures (including placebo) were greater than baseline values.” J.A. at 481 (emphasis added). “Only the [PPA]-cajfeine combination caused a peak [SBP] increase that was significantly greater than that after placebo.” J.A. at 480 (emphasis added). Peak DBP changes were significantly higher for all preparations than for the placebo. J.A. at 481.
E
Finally, the fifth paper, again using the same 16 subjects, tested 150 mg PPA, 75 mg PPA with 400 mg caffeine, and simply 75 mg PPA. It produced some interesting quota*985tions that the court relies on. The court states that “all three subject groups experienced three times more blood pressure rises than did the placebo group.” At 974 (emphasis added). This apparently is the court’s interpretation of a result that applied only to the peak increases in supine blood pressure, and refers to the amount of that peak increase, not how often the increase occurred.
Thus, the data in Table I, J.A. at 467, relating to the peak change in supine blood pressure, show increases as follows:
PLACEBO 150 PPA 75 + CAFF 75 PPA
SYSTOLIC 13 47 30 24
DIASTOLIC 9 32 25 27
However, the article interprets these numbers to mean that three hours after ingestion (previously thought to be a peak time) “only after 150 mg were the differences significant.” J.A. at 465-66 (emphasis added).
It is significant, moreover, that the figures for standing blood pressure showed much less difference, with average values three hours after ingestion showing the following results:
PLACEBO 111/76
150 PPA 116/74
75 PPA 121/81
75 + CAFF 111/79
Table II, J.A. at 468. These results show little systematic difference, and in fact show the anomalous result that 150 mg had less effect than 75 mg, and that 75 mg with caffeine had virtually no effect at all. Such results, comprising a random and erratic pattern, might well have been expected from a study using such small numbers and producing such small effects.
Finally, the article found that, when standing blood pressure was taken again, at the critical three-hour point, after the subjects stood in place for 5 minutes, “all blood pressures were similar across all four drugs [includes placebo ] and similar to baseline [pre-ingestion] values.” J.A. at 466.
Nevertheless, the court picks up the authors’ language that “some individual peak blood pressures that were reached were worrisome _” At 974 (quoting J.A. at 470-71). However, that conclusion rests on single maximum values that are, almost by definition, anomalous with such a small number of subjects. The full report of the data shows the following absolute maximum supine blood pressure values (the highest SBP for any individual and the highest DBP for any individual, regardless of whether the two recorded amounts were in the same individual and regardless of the pre-testing values for that individual):
PLACEBO 172/118
150 PPA 230/140
75 + CAFF 178/120
75 PPA 210/160
Table I, J.A. at 467. Again, this examination of single high values shows that much of what is going on is simple random fluctuation. The values for 75 mg alone are much higher than for 75 mg plus caffeine, even though all solid studies show that the addition of caffeine is more effective in raising blood pressure. The DBP reading for 150 mg is lower than the reading for 75 mg. Also, the 75 mg plus caffeine formulation shows essentially the same result as the placebo alone, further undermining any relevance of these erratic numbers.
It is very significant that the authors’ own “Disoussion” section only makes the statement that “our data support previous studies which conclude that [PPA], in doses of only double or triple recommended, elevates supine (but not standing) blood pressure to clinically significant levels.... ” J.A. at 470 (emphasis added) (footnotes omitted).
Finally, the court quotes a short passage from this study that relates to a 1989 Pentel study not in the record, see J.A. at 471 & 474 n. 24,9 for the proposition that Pentel had two subjects with “some blood pressure su-persensitivity to recommended doses of [PPA].” Without the study being in the record, we have no way of knowing what numbers are implied by “supersensitivity.” We also do not know what is meant by “recommended” doses here, especially in light of the analysis of the 1985 Pentel study, see infra at 987, showing that the doses used in that study were much more powerful in *986effect than the 75 mg PPA that we are concerned with here.
F
Thus, the combined effect of all “five papers,” reporting on two studies, involving some or all of the same sixteen subjects in four of the papers and at most fourteen other subjects in the remaining paper, is fully consistent with the much larger studies published by Blackburn, Noble and Liebson: Some PPA probably elevates blood pressure a little, but not much. A lot of PPA, or PPA with caffeine, will raise blood pressure more, and possibly to levels with clinical implications. In fact, the fifth and final Zalo-ga/Lake paper faithfully records the actual facts: “[PPA] appears to cause some blood pressure elevation in humans.... An oral dose of 75-mg sustained-release [PPA] ... causes statistically significant but clinically trivial blood pressure increases in some studies_” (emphasis added). J.A. at 464. That is what the studies as a whole stand for, and that is not a sufficient basis for Zaloga’s opinion. Beyond that, no matter how artfully the authors worded the papers, or how deftly Dr. Zaloga or the court summarizes them,10 there is nothing but “data dredging” to show a minor effect here or a single anomalous value there.
IV
Next the court turns to Zaloga’s reliance on the 1985 Pentel study and the Horowitz study for some exciting quotations. The court cites Pentel as finding “blood pressure rising significantly [apparently meaning “amount,” not “statistical significance” — you cannot get statistical significance from “some” individuals] in some individuals after 75 mg doses” with a “greatest individual rise in blood pressure” of 55/35. At 974 & n. 8. Horowitz is cited as saying that “[s]ubjeets receiving 85 mg PPA suffered increases in blood pressure” and that “potentially dangerous rises in blood pressure may occur after ingestion of a single capsule of PPA-containing substances and that a report of cerebral hemorrhage after ingestion of 85 mg confirms that these adverse effects may be life-threatening.” At 974.
However, the court discounts the fact, reported and confirmed in Dr. Zaloga’s own papers, that these two studies used much more potent forms of PPA. Pentel used immediate-release rather than timed-release capsules in studying his six subjects, and Horowitz studied a substantially more potent non-racemic isomer11 of PPA that is not approved for marketing in the United States, rather than the racemic form used in this country.
A
Both the first and second Zaloga/Lake papers note that Horowitz used dosages that were more potent than the approved 75 mg of PPA, that were delivered in immediate-release form, and that were non-racemic. See J.A. at 291, 458. Remarkably, during his deposition, Zaloga vigorously denied that the Australian Trimolets compound used by Horowitz would translate into approximately the equivalent of 170 mg of PPA in sustained-release form. Zaloga challenged his questioner:
I would disagree with that. Unless someone can produce that information for me, I don’t believe it.
*987I don’t think you can just assume, because it was in a different form or different isomer form, that you can just add them up that way, so I disagree with that.
Depo.Tr. at 518-19. Zaloga’s contentiousness is surprising because his own co-authored statement, in the second paper cited by the court, specifically says that Horowitz’s single 85 mg non-racemic immediate-release capsule was equivalent to 170 mg of the racemic American preparation of PPA. J.A. at 458, 460. See also the third paper, J.A. at 298. Thus, Zaloga himself produced the information that Horowitz used an amount even greater than the greatest amount tested in any of the other studies, and above the value that all agree may produce clinical effects.
B
Pentel’s study of three men and three women also used an immediate-release form of PPA, see J.A. at 294 & 299 n. 21, 304, 464 & 478 n. 10, and is referred to in the second Zaloga/Lake paper as “inconclusive” as to PPA’s effects on blood pressure. J.A. at 458.12 The fifth Zaloga/Lake paper cites the Pentel study in noting that 75 mg of immediate-release PPA equates to 150 mg of sustained-release PPA. J.A. at 464.
Thus, neither the Horowitz nor the Pentel study shows anything about the kind of 75 mg PPA dose that would have been taken by Bryan Glaser.
V
The final support for Zaloga’s conclusion is proffered in the form of adverse case reports, variously reported as 140 to 142. Zaloga appears to be referring to some version of an article by Lake, et al., which is referred to in various of the papers in various ways:
* J.A. at 484 n. 16 “Adverse drug effects attributed to [PPA]: a review of 142 case reports. AmJMed [In press].”
* J.A. at 473 n. 6 “[Same]. AmJMed 1940 [sic]; 86:195-208.”
* J.A. at 299 n. 39 “Adverse drug effects attributed to [PPA]: a review of 140 case reports (submitted for publication).”
* Defendant’s brief at page 32 refers to the article as:
89 Amer.J.Med. 195 (August 1990).
Whatever the article’s actual citation, it is not in the record. However, Zaloga’s discussion at the deposition clearly shows that only 12 or 13 reported incidents involved 75 mg of PPA (and we do not know whether any of these were immediate-release or non-race-mic), and he admitted that “some” of those 12 or 13 cases (we do not know how many) involved “combinations with other drugs.” J.A. at 391. The only individual cases discernible from the record are the three quoted at 972-73 n. 1 of the court’s slip opinion (quoting J.A. at 390). Without more, and without the original reports, these are the rankest type of anecdotal data.
In commenting on regulations of the federal Food and Drug Administration (FDA) that state that the agency will not consider “[i]so-lated case reports, random experience, and reports lacking the details which permit scientific evaluation,” 21 C.F.R. § 314.126(e), the Supreme Court has observed that the FDA’s exclusion of such anecdotal evidence is amply justified by legislative history that shows “a marked concern that impressions or beliefs of physicians, no matter how fervently held, are treacherous.” Weinberger v. Hynson, Westcott & Dunning, Inc., 412 U.S. 609, 619 & n. 14, 93 S.Ct. 2469, 2478 & n. 14, 37 L.Ed.2d 207 (1973); E.R. Squibb & Sons, Inc. v. Bowen, 870 F.2d 678, 685 (D.C.Cir.1989); see also Hagaman v. Merrell Dow Pharmaceuticals, Inc., 1987 U.S.Dist. LEXIS 6124 at *22-24 (D.Kan. June 26, 1987) (discussing reasons that anecdotal drug experience reports are inherently unreliable, unscientific, regarded by the FDA as unhelpful, and difficult for courts to verify). In this *988ease, these three tersely stated anecdotal events certainly cannot support any kind of conclusion as to causation.
VI
The above demolition of the interpretation that Zaloga and the court put on the studies would be ample to support the summary judgment granted here. However, there is more. In casually quantifying the competing “probabilities” as to what may have caused Glaser’s fall, Zaloga’s numbers are fatally flawed by a diagnostic technique that depends almost wholly on his ignoring the possibility that Glaser fainted. On numerous occasions, Zaloga discusses the case as though we have no reason to believe that Glaser was anything other than a “very athletic person” in the peak of health. Despite the contrary evidence, he disdains any suggestion that Glaser’s general state of health at the time was other than optimal:
[I]t was my opinion that in a young, healthy individual who had no underlying disease that would have predisposed him to an intercerebral bleed and came into the emergency room with severe hypertension and with a history of taking a compound which is known to cause hypertension and intercerebral bleeds, my personal opinion was that in fact that would be the leading diagnosis in this case.
J.A. at 353.
So I’m left with a situation of having a relatively young, healthy individual without underlying predisposed disease who suddenly has an intercerebral hemorrhage. In my mind, in that case the most — the thing that I would — would look towards would be use of some kind of a drug that could cause hypertension.
J.A. at 362.
Also I don’t have any evidence presented to me that suggests that he slipped or he fell. He wasn’t in line for a prolonged period of time to have le[]d to pooling.
We have no evidence that he was severely dehydrated or other reason to have fallen.
J.A. at 365.13
People just don’t walk in and fall and have this kind of a bleed.
J.A. at 387.
People can faint anywhere, but lots of people are fatigued and never faint as well.
So I have no reason to understand why he would have fainted from being fatigued. He had been fatigued for a long period of time and hadn’t fainted.
J.A. at 388.
In short, Zaloga almost always emphasizes that strokes with no apparent reason or precipitating trauma in young, healthy people are rare, and in that case one should look to a drug as a cause. I don’t quarrel with that. However, his logic for dismissing as an “apparent reason” a faint due to prolonged standing while wearing a coat inside a hot, crowded room, following weakness from flu and stringent dieting, is without foundation. He is simply wrong about the amount of standing. He has no basis for his comments regarding the temperature conditions in the room at the time. The notion that fainting can be ruled out because many people don’t faint, or because this person didn’t faint previously, is unsupported conjecture. Indeed, many people don’t have intercerebral bleeding. There is no demonstrable basis for the 80%-15% assignment of “probabilities” between PPA and fainting as the cause of bleeding. Rather, the numbers are an extension of the witness’s unsupported proffer of personal opinion, unrelated to the scientific method and contrary to the evidence.
YII
A good bit of the disagreement that I have with the court’s opinion is in its use, or acceptance of Dr. Zaloga’s use, of words such as “cause” and “significantly.” In many cases, “cause” is used when what is actually meant is that some effect is associated with, *989or follows, an earlier stimulus. Thus, Zaloga places enormous weight on a single blood pressure reading of 210 systolic or 160 diastolic after taking PPA. See the court’s opinion, at 974. However, as discussed earlier, supra at 985, in the very same study the taking of a placebo was followed by a blood pressure reading of 172/118, well above the hypertension level. Is it sensible to say that this hypertension also was “caused” by the placebo? The short answer is no. The whole point of trials of significant size, and the emphasis on means and standard deviations, is to avoid the temptation to make judgments based on anomalous data.
Second, Zaloga makes no attempt to show that the “extreme” values that he reports and emphasizes are anything other than what results from the normal application of the reporting of a mean and a standard deviation. In other words, he does not assert nor attempt to demonstrate that 75 mg of PPA is peculiarly likely to produce extreme values. Finally, even if there were an attempt to demonstrate that 75 mg of PPA could, in some person, somewhere, under some conditions, “cause” an extremely high blood pressure reading, there is no attempt to tie that particular susceptibility to this plaintiff, Bryan Glaser. Again, taking Dr. Zaloga’s testimony and this opinion at face value, there would be no reason that any person who faints and suffers damage after taking one of the 16 billion doses of PPA consumed annually in the United States, should not be able to bring his claim to a jury. I do not believe such a result is compatible with the law in our circuit or the decisions of the Supreme Court. I therefore dissent.
. A typical cup of coffee contains approximately 100 mg of caffeine. See Rudolf Noble, "A Controlled Clinical Trial of the Cardiovascular and Psychological Effects of Phenylpropanolamine and Caffeine,” 22 Drug Intelligence & Clinical Pharmacy 296, 298 (1988), reprinted in Joint Appendix at 318.
. Dr. Zaloga's fee for consulting in this case has been $200 per hour. Depo.Tr. at 20.
. The teller, Jaqueline Kulchycki, had been deposed on June 24, 1992. J.A. at 426. While he was being deposed on July 27, 1992, Zaloga confirmed that he had read the transcript of Kulchycki's deposition. Depo.Tr. at 36.
. Dr. Zaloga did not explain how, without "any evidence presented to me that suggests that he slipped or he fell," he assigned a 15% probability to such an event.
. "I think it's possible — I—I mean, one — two [doses] would be more likely, but I think the possibility — if he had two [doses], I would say his — his probability is much higher — say, 90 percent. Having ingested one [dose], my idea of his probability would be somewhat lower — at, say, 80 percent.” Depo.Tr. at 203-04 (emphasis added).
. Although Dr. Bodzin did not determine what caused the fall, he did conclude that the bleeding in Glaser’s brain was caused by the trauma to his head upon falling. Bodzin Depo.Tr. at 31, reprinted in J.A. at 199.
. It should be noted that a study can report “blood pressure” in at least 8 different ways, depending on whether the systolic (first or higher number) or the diastolic (second or lower number) is considered, then whether the subject is "supine" (lying down) or standing, and whether mean or maximum values for a group are reported. This allows a great deal of picking and choosing for a data analyst witness. See, e.g., Liebson, J.A. at 321 ff.
. As to the second paper, at 973 n. 3, which reported on a sample of fourteen, it is not clear who these subjects were, but Zaloga stated that his group conducted only two studies. J.A. at 355.
. This is not the Pentel study referred to in the court’s opinion, at 974 n. 8, which is a 1985 study. See infra § IVB.
. At one point during his deposition, Zaloga attempted to distinguish his papers from the Blackburn, Noble, and Liebson studies by saying that “they never did toxicity studies. I believe that 75 milligrams [of PPA] itself has toxicity to it in a small number of patients. In my mind, doubling the dose would create toxicity in the majority of patients. And I’m going to just say I think it would cause it in close to a hundred percent, because our studies suggest every patient who we gave 150 milligrams to developed unacceptable hypertension, in my opinion." Depo.Tr. at 228 (emphasis added). The court cites this testimony. At 975 n. 11.
Zaloga does not indicate what he means by “toxicity studies,” nor does he indicate that he performed any such studies that the other authors did not.
. In Dr. Zaloga’s words, an isomer is an analog whose bonds can be shifted in two directions called "L” and "D.” Depo.Tr. at 157. Although he says that he knows of no published evidence that the Australian non-racemic isomer has more activity than does the racemic isomer that is approved for sale in the United States, his own co-authored studies indicate otherwise.
. In his deposition, Zaloga now suggests that sustained-release doses of PPA can cause greater elevations in blood pressure than immediate-release doses, Depo.Tr. at 130-32, just as he now claims that the Horowitz non-racemic isomers produce effects that are indistinguishable from racemic isomers. See supra § IV A. However, as with the Horowitz isomers, Zaloga’s own coauthored studies contradict the thrust of his deposition testimony. Therefore, the fact that he is "clearly aware of the differences between the two compounds,” as the court notes, at 974 n. 9, does not lend greater credence to his testimony.
. It will be recalled that the teller testified that Glaser had been standing in a line with a 20-25 minute wait, wearing a coat inside a hot, stuffy, crowded bank lobby after coming in from a January day. J.A. at 427-28. He came to the bank immediately after being diagnosed with post-viral syndrome and an immediate history of flu-like symptoms, during a time when he was eating poorly. J.A. at 154-64.