NOONAN, Circuit Judge.

Margo Lynch, the minor daughter of Dennis and Margaret Lynch, and her parents, all citizens of Massachusetts, sue Merrell-National Laboratories, Division of Richardson-Merrell, Inc., an Ohio corporation. Jurisdiction is based on diversity of citizenship; Massachusetts law applies. The case is one of a number which have arisen in which the drug Bendectin is alleged to have caused a birth defect. The district court gave judgment for the defendant. Lynch v. Merrell-National Laboratories, 646 F. Supp. 856 (D.Mass. 1986). We affirm the district court.

FACTS

Bendectin, as originally designed, was a combination of three components: dicyclomine hydrochloride, an antispasmodic drug, earlier marketed under the name Bentyl; doxylamine succinate, an antihistamine, earlier marketed under the name Decapryn; and pyridoxine (Vitamin B6). The composite was approved by the Food and Drug Administration for sale in the United States in 1956 for use in alleviating morning sickness in pregnancy. In 1976 the composition was altered to omit Bentyl. The defendant was its manufacturer. In 1983, in the face of a host of lawsuits and some congressional criticism, the defendant withdrew it from the market.

In the fourth week of her pregnancy, in the last week of June 1974, Margaret Lynch began experiencing severe 24-hour-a-day morning sickness. Her physician, Ambler Garnett, Jr. of Newburyport, prescribed Bendectin, which she then took for the remainder of her term.

On February 17, 1975, Margaret's daughter Margo was born. She was born without a right hand and without the lower portion of her right forearm. She needed special medical care, which her parents provided. Her losses were, of course, permanent. She and her parents suffered from the cruel injury.

William J. Driscoll III, the obstetrician-gynecologist who delivered Margo, had no opinion as to the cause. Eventually, however, the Lynches came to the conclusion that Bendectin was the cause. On February 10, 1985 they brought suit against its maker, alleging that the manufacturer had negligently designed, developed, selected materials for, manufactured, assembled, tested, inspected, advertised, sold, promoted and distributed the drug; had negligently failed to warn its users; had falsely and misleadingly advertised it; had breached both express and implied warranties of safety; and had put on the market a drug dangerous and defective in design. They sought total damages of $7.5 million.

PROCEEDINGS

The Lynches' case was transferred to the Southern District of Ohio where since 1982 similar cases had been brought for pre-trial proceedings under Judge Carl Rubin. Over 1,000 cases were consolidated for this purpose under the authority of 28 U.S.C. § 1407. The Lynches stipulated to be bound by the pretrial discovery in these proceedings. Given the option, however, of joining in the trial or having their case remanded to the district court in Massachusetts, the Lynches elected the latter course.

The consolidated cases proceeded to trial before Judge Rubin in Cincinnati. Just as trial began, he certified the plaintiffs as a class for settlement; the certification was reversed by mandamus. In re Bendectin Products Liability Litigation, 749 F.2d 300 (6th Cir. 1984). After a trial on the merits, the jury found for the defendant, Merrell-National Laboratories. In re Richardson-Merrell, Inc. "Bendectin" Products Liability Litigation, 624 F. Supp. 1212 (S.D.Ohio 1985), appeal pending.

In Boston in 1986 the defendant moved for summary judgment against the Lynches. The district court granted the motion, ruling that the Lynches were collaterally estopped by the judgment in the multi-district litigation and that in any event they had presented insufficient proof of causation to permit a reasonable trier of fact to conclude that Bendectin had caused the birth defect of Margo Lynch. The second part of the district court's decision depended on its ruling that the expert testimony proffered by the plaintiffs was inadmissible. The Lynches appealed.

ISSUES

I. Were the Lynches collaterally estopped by the judgment in the multi-district litigation?

II. Was the proffered expert testimony admissible?

III. Was there sufficient evidence of causation to avoid summary judgment?

ANALYSIS

I. Collateral estoppel. The question is close and the thoughtful analysis by the district court all but compelling. A preliminary question is whether, in litigation based on diversity of citizenship, the doctrine of estoppel to be applied is federal or state. See Burbank, "Interjurisdictional Preclusion, Full Faith and Credit and Federal Common Law: A General Approach" 71 Corn.L.R. 733 (1986); Degnaon, "Federalized Res Judicata," 85 Yale L.J. 741 (1976). The parties and the district court have assumed that the law is federal. But the teaching of Erie R. Co. v. Tompkins, 304 U.S. 64, 78, 58 S.Ct. 817, 822, 82 L.Ed. 1188 (1938) is clear as to the law to be applied in diversity:

Except in matters governed by the Federal Constitution or by Acts of Congress, the law to be applied in any case is the law of the State . . . There is no federal general common law.

The law on collateral estoppel is as substantive in its effect as the law of contracts or torts. It is not a law created by the Constitution or an Act of Congress. This court has applied Massachusetts law on collateral estoppel in a case where jurisdiction rested on diversity and an earlier case, relied on by the defendant, had been decided in a Massachusetts court. Standard Accident Ins. Co. v. Doiron, 170 F.2d 206, 207 (1st Cir. 1948). On the other hand, the argument must be considered that the federal system has the power and duty to protect its own judgments, and that when the effect of a federal judgment is at issue, even in a diversity action, federal law must govern. Kern v. Hettinger, 303 F.2d 333, 340 (2nd Cir. 1962); cf. Byrd v. Blue Ridge Cooperative, 356 U.S. 525, 537, 78 S.Ct. 893, 900, 2 L.Ed.2d 953 (1958).

We need not resolve the question for every possible context. In this case involving the effect of multi-district litigation in a federal court, we believe the same result would follow under either federal or state law. Massachusetts has abandoned the old requirement of strict mutuality and permits a defendant to bind a plaintiff by the plaintiff's previous defeat in another case presenting the same issue, Home Owners Federal Savings and Loan Ass'n v. Northwestern Fire and Marine Insurance Co., 354 Mass. 448, 238 N.E.2d 55 (1968); and Massachusetts permits a plaintiff to bind a defendant by a previous criminal conviction. Aetna Casualty and Surety Co. v. Niziolek, 395 Mass. 737, 481 N.E.2d 1356, 1361 (1985). But so far Massachusetts has not permitted a defendant to bind a plaintiff by the defendant's previous victory in another case presenting the same issue. Federally, the traditional requirement of mutuality has been eliminated to permit a defendant to invoke estoppel against a plaintiff who lost on the same issue to an earlier defendant, Blonder-Tongue Laboratories, Inc. v. University of Illinois Foundation, 402 U.S. 313, 91 S.Ct. 1434, 28 L.Ed.2d 788 (1971), and to permit a plaintiff to invoke estoppel against a defendant who lost to a prior plaintiff. Parklane Hosiery Co. v. Shore, 439 U.S. 322, 99 S.Ct. 645, 58 L.Ed.2d 552 (1979).

The preference of both state and federal decisions for judicial economy might well be read as the district court read the federal cases, to indicate that mutuality should also be dispensed with where, as here, a plaintiff presents precisely the issue on which the defendant has prevailed against a prior plaintiff. Such an extension of estoppel has additional appeal where, as here, the plaintiffs had ample opportunity to join the multi-district litigation and chose not to do so, perhaps believing that if the defendant lost in Cincinnati the plaintiffs would avail themselves of that victory while if the defendant won, the plaintiffs would have a second shot in Boston.

Nonetheless we do not believe that either federal or state law would warrant the application of collateral estoppel in the present context. In rejecting certification of the class, the Sixth Circuit noted the teaching of Parklane Hosiery, supra, 439 U.S. at 330, 99 S.Ct. at 651, that later plaintiffs could not invoke to their benefit a favorable result in mass tort litigation. In re Bendectin Products Liability Litigation, supra, 749 F.2d at 305. If later plaintiffs could not, why should the defendant? Judge Rubin himself, dealing with plaintiffs who had opted out of the multi-district litigation, refused to give estoppel effect to the ultimate result in favor of the defendant. In re Bendectin Cases, No. 85-0996 (E.D.Mich. 1986). Estoppel should not be applied unless the plaintiffs had a fair and full opportunity to litigate. Parklane Hosiery, supra, at 327, n. 7, 99 S.Ct. at 649, n. 7, citing Blonder-Tongue at 329, 91 S.Ct. at 1443. But the plaintiffs were allowed to think that they could withdraw from Cincinnati and lose nothing. They did not have a fair opportunity when they understood that their withdrawal would not prejudice them. If they were now bound, the multi-district litigation would in effect have been a class action leaving the Lynches no true option. We believe that their freedom to withdraw and come back to Boston was not illusory.

II. The district court, addressing the admissibility of the plaintiffs' experts' testimony, carefully reviewed the records of testimony in other Bendectin trials, expert depositions and affidavits in this case, and scientific studies since 1963 exploring the relation of Bendectin to birth defects. Lynch v. Merrell-National Laboratories, 646 F. Supp. 856, 862-866. We incorporate this review by reference and sum up the main conclusions, supplemented by our own review and analysis:

1. Human development in the womb carries a substantial risk of death or abnormality. Disasters of deformation, fatal fetal accidents occur. Fifteen percent of all clinically recognized pregnancies result without human intervention in death before birth. Three to six percent of children born have malformations. Beckman and Brent, Mechanism of Known Environmental Teratogens: Drugs and Chemicals, 13 Clinics in Perinatology 649, 652 (1986).

2. For most of these disasters there is no one to blame. The cause of sixty-five percent of malformations observable in the first year of life is unknown. Another twenty to twenty-five percent are due to genetic factors. Seven percent are due to maternal conditions such as rubella, syphilis, nutritional deficiencies, and narcotics addiction. One to two percent are due to mechanical difficulties in delivery. The known impact of "chemicals, drugs, radiation, and hyperthermia" is a causal factor in less than one percent of the deformations. These estimates come from impartial pediatricians. Id., 650. The plaintiffs' own witness, Alan Kimball Done, had a very similar opinion. According to him, the cause of 75 to 80 percent of all birth defects is unknown. Lynch at 863.

3. According to Murray G. Feingold, Physician-in-Chief at the National Birth Defects Center, deposed by the plaintiff, the incidence of limb reduction in the population at large is less than 1 per 1,000 live births. Limb reductions, then, are relatively rare and so are particularly shocking. They are a subspecies of the larger class of all birth defects whose incidence is more common. In ninety-five percent of the cases of limb reduction, Doctors Done and Feingold agree, the cause is unknown.

4. After the Thalidomide-caused catastrophe of the 1960's, popular suspicion found it easy to link any birth defect with a drug taken during pregnancy. Limb reduction, in particular, acted as a red flag. Wilson, Misinformation About Risks of Congenital Anomalies, in Marois, Prevention of Physical and Mental Congenital Defects: Part C (1985) 165, 167. Under the name Debendox in Australia and Britain and under the name Lenotan in Germany, Bendectin — widely used in pregnancy — was an object of such suspicion, as it was in the United States. Stimulated by the suspicion, extensive epidemiological studies were undertaken. Researchers examined Bendectin in relation to birth defects and attempted to determine if there was a correlation between the use of Bendectin and the defects from which a conclusion of causation might be drawn. These studies were conducted by Australians, Americans, Englishmen, and Germans. The researchers were specialists in a variety of fields from gynecology, public health, embryology, and teratology, to epidemiology proper. The researchers had every motive to detect a malignant drug inflicting horrible injuries. The first studies began in 1963; studies are still being published in 1986. No correlation, much less a causal connection, has been demonstrated between the use of Bendectin and limb reduction. Beckman and Brent, supra at 653.

5. A study by Dr. Steven H. Lamm shows that, while Bendectin usage declined from 1 million new therapy starts in 1979 to zero in 1984, there has been no change in the incidence of birth defects. Lynch at 864. Lamm holds an M.D. from the School of Medicine of the University of Southern California, 1956, and a diploma in Tropical Public Health from the London School of Hygiene and Tropical Health, 1974. He is the president of Consultants in Epidemiology and Occupational Health, Inc. and first became involved in the Bendectin litigation when asked to prepare a report by the guardian ad litem for a plaintiff in the multi-district litigation. This report reflected this non-relation between the use of Bendectin and limb reduction.

6. On the basis of the epidemiological evidence to date, Bendectin is as likely as aspirin to cause limb reduction. Because many mothers take aspirin during pregnancy, as they once took Bendectin, there will always be some children whose mothers took aspirin who were then born with defects including limb reduction. The coexistence of the defect and the taking of aspirin does not prove that one caused the other. Rather, the connection is one of "happenstance." Wilson, Misinformation About Risks of Congenital Abnormalities, supra, at 167. To blame aspirin or to blame Bendectin would be an "aberration of reason." Id. at 168. The association of Bendectin with limb reduction is in the view of the health-care community an instance of popular delusion and error. Beckman and Brent, supra at 653.

We face then a situation in which limb reductions are a fairly unusual subspecies of defect, in which the origin of most limb reduction is unknown, in which world-wide scientific investigations of Bendectin have produced no evidence establishing that Bendectin causes limb reduction, and in which the irrelevance of Bendectin to the incidence of limb defects has been demonstrated. The ignorance that prevails as to the etiology of most birth defects does not mean causation in a given case could not be proven; it does mean that there is a large terra incognita where gossip and guesswork abound, so that courts must carefully control the basis for testimony pointing to a particular cause. A new study coming to a different conclusion and challenging the consensus would be admissible evidence. Without such a study there is nothing on which expert opinion on Bendectin as a cause may be based. The plaintiffs offered no new study.

The plaintiffs did offer the opinion of Dr. Done, a 1952 graduate of the medical school of the University of Utah. Done was an assistant professor of pediatrics at Stanford from 1958 to 1960; a professor of pediatrics at Utah from 1960 to 1971; a special assistant to the director of the Bureau of Drugs in the Food and Drug Administration, 1971-1974; and a professor of pediatrics and pharmacology at the College of Medicine, Wayne State University, 1975 to date. Done had come to the conclusion that Bendectin could cause limb reduction on the basis of in vivo animal studies, in vitro animal studies, and the study of "analogous" chemicals. The district court for the reasons stated at 646 F. Supp. 865-866 rightly found that these studies did not furnish a foundation for Done's opinions. Studies of this sort, singly or in combination, do not have the capability of proving causation in human beings in the absence of any confirmatory epidemiological data. See In re "Agent Orange" Product Liability Litigation, 611 F. Supp. 1223, 1241 (D.C.N.Y. 1985).

The plaintiffs also offered the opinion evidence of Shanna Helen Swan, a holder of a 1963 doctorate in statistics from the University of California at Berkeley. Swan had served from 1969 to 1975 as senior biostatistician in a Kaiser Health contraceptive drug study; been associate professor of mathematics at California State University, Sonoma, from 1974 to 1979; directed between 1979 and 1981 the training program in biostatistics and epidemiology at the School of Public Health of the University of California, Berkeley; and, while remaining in this school, has been since 1981 the chief of the Methodology and Analysis Unit, Epidemiology and Statistics, Department of Health Services of the State of California.

Swan's opinion as to causation was based upon her reanalysis of data collected between 1970 and 1978 by the Metropolitan Atlanta Congenital Defects Program. This data had been previously analyzed by four members of the Center for Disease Control. The subjects of the study were an entirely abnormal set — 1,231 children with birth defects and their mothers. Nine and one half percent of the mothers reported using Bendectin. The rate of limb reduction in their defective children was compared to the rate of limb reduction in the rest of the group. This abnormal group was used as a control in order to wash out "maternal-recall bias", because mothers of defective children tend to recall drug usage more readily than mothers of normal children. Comparing mothers who all had a reason for recall, the researchers found "weak" but not causal relationship between the use of Bendectin and amniotic bands and concluded that the study "excluded" a risk of limb reduction from the use of Bendectin. Cordero, Oakley, Greenberg, and James, Is Bendectin a Teratogen?, 245 Journal of the American Medical Association 2307, 2310 (1981).

Swan testified in the multi-district litigation in Cincinnati, and the district court here took note of her testimony as it was excerpted by the defendant. According to this testimony, Swan chose the group for re-analysis because it was a federal study, a large study, and there were "no normal controls." She then re-read the interview forms of the 1,231 mothers, looking for any misclassifications as to the drug used or dates of exposure. She chose as her control group not the entire universe of non-Bendectin users but the children afflicted with Down's syndrome and other genetic disorders. She chose them because such disorders were the only defects which a priori could not be associated with Bendectin. Comparing the Bendectin-users' children to the genetically-disadvantaged children, she concluded that there was a risk of limb reduction associated with Bendectin, and she was apparently willing to testify that Bendectin in fact causes limb reduction.

What Swan might have testified to is only apparent because in opposition to the motion for summary judgment the plaintiffs did not file any affidavit, deposition or trial testimony by her; and the only evidence of her opinion is the excerpt of her testimony produced by the defendant. In that excerpt Swan acknowledges that relative to any other subgroup in the 1,231 sample "the genetic groups" had a lower risk of 0.57 for certain categories of birth defect — that is, they had almost forty percent better chance of not having such defects. As far as appears from what is in the record, Swan made no allowance for the possibility that the very fact of having such a severe genetic deficiency as Down's syndrome might operate to make other rare deficiencies such as limb reduction less likely to occur in the control group — that is, that the combination of Down's syndrome and another major misfortune might be extremely unusual. Without accounting for this possible skewing of the control group, Swan's basis for her comparative conclusion is not apparent.

Swan's study has never been refereed or published in a scientific journal or elsewhere. We are informed of it only by the defendant's excerpts. On the basis of what we have, it could not form the foundation for an expert opinion challenging the scientific consensus and making the issue of causation a factual question to be decided by the jury.

Our review would not be complete without consideration of cases in which jurors and judges — all reasonable men and women — have concluded that Bendectin is in fact a teratogen and have awarded damages against the defendant. The leading case is Oxendine v. Merrell-Dow Pharmaceuticals, 506 A.2d 1100 (D.C.App. 1986). Here a jury awarded $750,000 to a child born with a shortened right forearm and a three-fingered right hand. The trial judge entered judgment notwithstanding the verdict and in the alternative ordered a new trial. The appeals court reversed and sent the case back for trial only on the issue of punitive damages.

In that case both Done and Swan were permitted to testify, so that, as the district court here observed, the issue before the appeals court was not the admissibility of their testimony under the Federal Rules of Evidence, but the strength of the case before the jury. Lynch at 862, n. 3. The precedential force of Oxendine, however, cannot so easily be escaped. Swan's testimony was given in rebuttal and, so far as appears from Oxendine at 1109, was a criticism of selected epidemiological studies relied on by the defendant and did not show causation by Bendectin. Done's testimony is more significant: it went to causation. Done not only testified to the animal studies in vivo, the animal studies in vitro, and the chemical studies which made him suspect Bendectin as teratogen; he also used epidemiological evidence by making his own re-analysis of a study previously done which had concluded to the absence of a causal relation. This study, published in 1963 by two employees of the defendant, studied 2,218 pairs, one of them using Bendectin in the first trimester, the other not; it found 21 congenital abnormalities in the control group, 11 in the Bendectin group and concluded that the difference was insignificant. Bunde and Bowles, A Technique For Controlled Survey of Case Records, 3 Current Therapeutic Research 245, 248 (1963). Done excluded various pairs used in the study and ended with the conclusion that the Bendectin users had a 30 to 80 percent greater likelihood of bearing a defective child than the non-users in the study.

The question that we must resolve is whether Done's reanalysis would have been an adequate epidemiological foundation for his expert opinion here. The district court, informed as to Oxendine, must have known of the re-analysis. No flaw in it was noted by the Oxendine appeals court, and none is pointed to by the defendant here. On the other hand, the plaintiffs in opposing the motion for summary judgment, did not provide an affidavit, deposition, or testimony from Done. His reanalysis, so far as it appears, has never been refereed or published in any scientific journal or elsewhere. The district court was, and we are, ignorant of its methodology and how Done went about eliminating the pairs, except that we know from Oxendine that he eliminated the Canadian pairs — how many we do not know. We are unable to say that his study, known only by its elliptical description by the appeals court in Oxendine, was sufficient foundation for Done's expert testimony.

We are referred to no appellate decision other than Oxendine, and in Oxendine at this time a petition for rehearing is pending. A verdict for the plaintiff has been given in Blum v. Richardson-Merrell, Inc., No. 82-1027 (Pa.Super., January 1987); and the jury disagreed and the judge refused to give a directed verdict for the defendant in Lanzilotti v. Merrell National Laboratories, Inc., No. 82-10183 (E.D.Penn., July 9, 1986). A jury has given compensatory damages of $20 million and awarded punitive damages of $75 million to a boy born with deformed arms and hands; the district court has let the compensatory damages stand and taken the punitive damages under advisement. Ealy v. Merrell-Dow Pharmaceuticals, No. 83-3504 D.D.C., July 14, 1987).

We are uninformed as to the evidence in any of these cases and therefore both abstain from criticizing them and deny that they show that factual issues existed on this record. The sight of a helpless mutilated youngster may evoke emotion along with the corresponding wish to make somebody pay for his or her plight. Judge Rubin has observed that the presence of handicapped youngsters could render a jury "unable to arrive at an unbiased judgment." In re Richardson-Merrell, supra, 624 F. Supp. at 1224. With this very real possibility of runaway emotion overcoming judgment, the district court's firm rejection here of foundationless expert testimony was necessary, admirable, and entirely within the discretion of the court under Federal Rules of Evidence 403 and 703. Ricciardi v. Children's Hospital Medical Center, 811 F.2d 18, 25 (1st Cir. 1987).

III. Under long-established Massachusetts law, causation in tort must be proved by a preponderance of the evidence, that is, the proof must be such as to make the defendant's causality "appear more likely or probable in the sense that actual belief in its truth exists in the mind or minds of the tribunal notwithstanding any doubts that may still linger there." Smith v. Rapid Transit, Inc., 317 Mass. 469, 58 N.E.2d 754 (1945) (per Spalding, J., quoting Lummus, J. in Sargent v. Massachusetts Accident Co., 307 Mass. 246, 250, 29 N.E.2d 825, 827 (1940)). That chances "somewhat favor" the defendant being the cause is not preponderant evidence. Id., 58 N.E.2d at 755. Much less would speculation as to possibilities be preponderant evidence. The defendant wins if the plaintiffs fail to show "that there was a greater likelihood or probability that the harm complained of was due to causes for which the defendant was responsible than from any other cause." Corsetti v. Stone Co., 396 Mass. 1, 483 N.E.2d 793, 805 (1985); Beaver v. Costin, 352 Mass. 624, 227 N.E.2d 344 (1967) (per Reardon, J.). The plaintiffs, even when the evidence is considered in the light most favorable to them, failed to meet this burden. Summary judgment for the defendant must be AFFIRMED.