United States Court of Appeals
`For the Eighth Circuit
`___________________________
`
`No. 19-2899
`___________________________
`
`In re: Bair Hugger Forced Air Warming Devices Products Liability Litigation
`
`------------------------------
`
`George Amador
`
` Plaintiff - Appellant
`
`v.
`
`3M Company; Arizant Healthcare, Inc.
`
` Defendants - Appellees
`____________
`
`Appeal from United States District Court
`for the District of Minnesota
`____________
`
`Submitted: March 16, 2021
` Filed: August 16, 2021
`____________
`
`
`Before GRUENDER, KELLY, and GRASZ, Circuit Judges.
`____________
`
`
`GRUENDER, Circuit Judge.
`
`In December 2015, the Judicial Panel on Multidistrict Litigation created and
`
`centralized the In re Bair Hugger Forced Air Warming Devices Products Liability
`Litigation (“MDL”) in the District of Minnesota (“MDL court”) for coordinated
`
`

`

`pretrial proceedings. Plaintiffs1 in the MDL have brought claims against 3M
`Company and its now-defunct, wholly owned subsidiary Arizant Healthcare, Inc.
`(collectively, “3M”). Plaintiffs assert that they contracted periprosthetic joint
`infections (“PJIs”) due to the use of 3M’s Bair Hugger, a convective (or “forced-
`air”) patient-warming device, during their orthopedic-implant surgeries. In July
`2019, on 3M’s motion, the MDL court excluded Plaintiffs’ general-causation
`medical experts as well as one of their engineering experts, and it then granted 3M
`summary judgment as to all of Plaintiffs’ claims. Subsequently, the MDL court
`entered an MDL-wide final judgment.
`
`
`Plaintiffs appeal. First, they argue that the MDL court abused its discretion
`in excluding their general-causation medical experts and engineering expert.
`Second, they argue that the MDL court erred in granting 3M summary judgment
`whether or not those experts were properly excluded. Third, they argue that the
`MDL court abused its discretion in denying Plaintiffs’ request for certain discovery.
`And fourth, they argue that the MDL court abused its discretion in ordering certain
`filings on its docket to remain sealed. Additionally, on appeal, Plaintiffs ask us to
`unseal those parts of the appellate record that duplicate the filings whose sealing on
`the MDL court’s docket they challenge.
`
`We reverse in full the exclusion of Plaintiffs’ general-causation medical
`experts and reverse in part the exclusion of their engineering expert. We reverse the
`grant of summary judgment in favor of 3M. We affirm the discovery order that
`Plaintiffs challenge. We affirm the MDL court’s decision to seal the filings Plaintiffs
`seek to have unsealed. And we deny Plaintiffs’ motion to unseal those same filings
`on our own docket.
`
`
`
`1Although George Amador is the captioned Plaintiff-Appellant, this appeal is
`
`brought by all Plaintiffs in the MDL to challenge several MDL-wide rulings.
`
`-2-
`
`

`

`I.
`
`
`In the mid-1980s, Dr. Scott Augustine invented the Bair Hugger, a forced-air
`
`device used to keep patients warm during surgery so as to stave off hypothermia-
`related complications that can arise during or after surgery. The device consists of
`a central heating unit, a hose, and a disposable perforated blanket that is placed over
`the patient. The central unit, which is often situated on or near the floor when in use,
`draws in air through a filter, warms that air (usually to a temperature significantly
`above the operating-room temperature), and blows it through the hose into the
`perforated blanket. The air exits the blanket through the perforations and keeps the
`patient warm. Typically, both the patient and the blanket are covered with surgical
`draping during operations, and the blanket is placed on a part of the body away from
`the surgical site, so the air does not blow directly onto the surgical site.
`
`Dr. Augustine marketed and sold the Bair Hugger through Augustine Medical,
`
`Inc., the company he founded and led as CEO until 2004. Around that time, Dr.
`Augustine was forced to leave Augustine Medical while under investigation for
`Medicare fraud. Augustine Medical then reorganized, and the division of the
`company that retained the Bair Hugger product line changed its name to Arizant
`Healthcare. In 2010, 3M acquired Arizant Healthcare and the Bair Hugger product
`line. Arizant Healthcare was dissolved in December 2014.
`
`After leaving Augustine Medical, Dr. Augustine developed the HotDog, a
`
`patient-warming device that transfers heat conductively to the patient by direct
`contact with the patient’s skin rather than by forced hot air. He then began a
`campaign to discredit his old invention and promote his new one. These efforts bore
`fruit. In March 2013, a plaintiff sued 3M and Arizant Healthcare in Texas state
`court, claiming that he contracted a PJI due to the Bair Hugger’s use in his hip-
`replacement surgery. Dr. Augustine worked with the law firm representing that
`plaintiff to prepare a “litigation guide” and solicitation letter for the purpose of
`fomenting more litigation against 3M. By December 2015, more than sixty
`materially similar cases against 3M had been filed in or removed to federal district
`
`-3-
`
`

`

`courts around the country. At that time, the Judicial Panel on Multidistrict Litigation
`ordered these cases centralized in the District of Minnesota for consolidated pretrial
`proceedings. See 28 U.S.C. § 1407(a). Nearly 6,000 lawsuits have since been filed
`as part of the MDL.
`
`In these cases, Plaintiffs allege that they suffered PJIs from the use of the Bair
`
`Hugger during their orthopedic-implant surgeries. PJIs are frequently caused by the
`introduction of microbes into the surgical site during surgery. Bacterial
`contamination is a particularly significant threat in orthopedic-implant surgeries
`because a PJI can be caused by very few microbes, possibly even a single bacterium.
`For this reason, it is standard for such surgeries to take place in “ultra-clean
`ventilation” operating rooms, where air is blown into the operating room through
`high-efficiency particulate air (“HEPA”) filtration at a uniform velocity. This
`HEPA-filtered “laminar” airflow blows over the patient, reducing the likelihood that
`operating-room airflow will carry ambient bacteria from nonsterile areas of the
`operating room into the surgical site.
`
`Plaintiffs advance two theories for how the Bair Hugger caused their PJIs
`
`during their orthopedic-implant surgeries. According to the “airflow disruption”
`theory, waste heat from the Bair Hugger creates convection currents that carry
`ambient bacteria from nonsterile areas of the operating room to the surgical site
`despite the laminar airflow, resulting in PJIs. According to the “dirty machine”
`theory, the Bair Hugger is internally contaminated with bacteria, which are blown
`through the blanket into the operating room, where they become ambient and
`eventually reach the surgical site, resulting in PJIs.
`
`
`In the master long-form complaint filed in the MDL, Plaintiffs asserted
`fourteen state-law claims against 3M, including negligence and strict liability (for
`failure to warn, defective design, and defective manufacture), among others.
`
`
`During discovery, Plaintiffs subpoenaed a third party, VitaHEAT Medical,
`LLC, to produce discovery regarding its “UB3,” a conductive patient-warming
`
`-4-
`
`

`

`device. Plaintiffs alleged that the UB3 was an alternative design to the Bair Hugger,
`making this discovery ostensibly relevant to their design-defect claims. See
`generally 63A Am. Jur. 2d Products Liability § 894 (May 2021 update) (“The
`existence of an alternative design may be used to establish that a product was
`unreasonably dangerous due to a design defect, and in some jurisdictions may be
`required.”). VitaHEAT objected on relevancy grounds, arguing that the UB3 was
`too different from the Bair Hugger to count as an “alternative design” for product-
`liability purposes. Plaintiffs then filed what they captioned a “motion to overrule”
`this relevancy objection. The MDL court denied this motion, agreeing that
`conductive patient-warming devices like the UB3 are too dissimilar from the Bair
`Hugger to qualify as “alternative designs,” meaning that this discovery was not
`relevant. Cf. United States v. One Assortment of 93 NFA Regulated Weapons, 897
`F.3d 961, 966 (8th Cir. 2018) (“The Federal Rules of Civil Procedure limit discovery
`to that which ‘is relevant to any party’s claim or defense . . . .’” (quoting Fed. R. Civ.
`P. 26(b)(1))).
`
`
`The parties jointly agreed to a protective order to limit the disclosure of
`confidential information that might be contained in filings entered on the MDL
`docket. Pursuant to this protective order, the parties submitted numerous filings
`under seal over the course of the litigation. As relevant to this appeal, 3M sought to
`keep seven such filings under seal over Plaintiffs’ objection, asserting that it would
`suffer competitive harm if any was unsealed. The MDL court agreed and ordered
`these files kept under seal.
`
`As the litigation progressed, 3M moved to exclude Plaintiffs’ general-
`
`causation medical experts (Dr. Jonathan M. Samet, an epidemiologist; Dr. William
`Jarvis, an infectious-disease specialist; and Dr. Michael J. Stonnington, an
`orthopedic surgeon) as well as Plaintiffs’ engineering experts (including Dr. Said
`Elghobashi and Michael Buck). 3M also filed a motion for summary judgment
`contingent on the exclusion of Plaintiffs’ general-causation medical experts. The
`MDL court denied in pertinent part the motion to exclude those experts and denied
`the motion for summary judgment.
`
`-5-
`
`

`

`Subsequently, Gareis v. 3M Co. became the first bellwether trial in the MDL.
`
`See generally 156 Am. Jur. Trials § 219 (May 2021 update) (explaining the
`bellwether-trial process in mass-tort litigation). Ruling on pretrial motions in
`Gareis, the MDL court excluded evidence of Plaintiffs’ dirty-machine theory. The
`case then proceeded to trial on the airflow-disruption theory, and Plaintiffs’ experts
`Dr. Jarvis, Dr. Stonnington, and Dr. Elghobashi (among others) testified. After an
`approximately two-week trial, the jury returned a verdict for 3M.
`
`After the Gareis trial, 3M moved for reconsideration of the MDL court’s
`orders refusing to exclude Plaintiffs’ general-causation medical experts as well as
`Dr. Elghobashi and denying 3M summary judgment. The MDL court granted 3M’s
`motion, excluding Plaintiffs’ general-causation medical experts as well as Dr.
`Elghobashi and granting 3M summary judgment on all claims. The MDL court then
`entered an MDL-wide final judgment. See In re Temporomandibular Joint (TMJ)
`Implants Prods. Liab. Litig., 97 F.3d 1050, 1055 (8th Cir. 1996) (“A transferee court
`has the authority to enter dispositive orders terminating cases consolidated under 28
`U.S.C. § 1407.”).
`
`
`Plaintiffs appeal, challenging the MDL court’s exclusion of their general-
`causation medical experts and Dr. Elghobashi, the MDL-wide grant of summary
`judgment in favor of 3M, the discovery ruling regarding conductive patient-warming
`devices, and the sealing of seven MDL docket entries. Plaintiffs also ask us on
`appeal to unseal on our own docket the records that they argue should have been
`unsealed on the MDL court’s docket.
`
`
`
`
`II.
`
`We begin by considering Plaintiffs’ challenge to the MDL court’s exclusion
`of their general-causation medical experts (Dr. Samet, Dr. Jarvis, and Dr.
`Stonnington) and engineering expert (Dr. Elghobashi). Dr. Samet, Dr. Jarvis, and
`Dr. Stonnington each offered general-causation opinions—that is, opinions that the
`Bair Hugger “is capable of causing the [PJIs] from which” Plaintiffs allegedly
`
`-6-
`
`

`

`suffered, see Junk v. Terminix Int’l Co., 628 F.3d 439, 450 (8th Cir. 2010)2—based
`on both the airflow-disruption theory and the dirty-machine theory. Dr. Elghobashi
`created a computational-fluid-dynamics (“CFD”) model to support the airflow-
`disruption theory. The MDL court generally treated the medical experts as a
`collective set (their opinions were essentially the same and were founded on much
`of the same evidence), and it excluded their opinions as unreliable because (1) it
`concluded there was “too great an analytical gap between the literature and the
`experts’ general causation opinions”; and (2) “the causal inferences made by the
`experts have not been generally accepted by the scientific community.”3 The MDL
`court also excluded Dr. Elghobashi’s model and opinion because (1) his conclusion
`about the Bair Hugger’s effects in real-world operating rooms relied on an unproven
`and untested premise, (2) there was too great an analytical gap between the results
`of his CFD and his conclusion about the Bair Hugger’s effects in real-world
`operating rooms, and (3) the CFD model was developed for litigation.
`
`For the following reasons, we reverse in full the exclusion of the medical
`experts’ opinions and reverse in part the exclusion of Dr. Elghobashi’s model and
`opinion. We first recite the principles that govern our analysis. We then analyze the
`reasons given by the MDL court for excluding the experts.
`
`2The opinions in question did not address specific causation—whether the
`
`Bair Hugger “in fact caused the harm from which” any particular MDL plaintiff
`suffered. See id.
`
`3The MDL court articulated a third reason; namely, that “the experts failed to
`
`consider obvious alternative explanations.” This reason applies specifically to the
`medical experts’ treatment of the epidemiological study on which they relied, see
`infra Section II.B.1, that found an association between forced-air warming and PJIs,
`see Federal Judicial Center, Reference Manual on Scientific Evidence at 597-600 (3d
`ed. 2011) (explaining that “[c]onsideration of alternative explanations” is one of nine
`factors that “guide epidemiologists in making judgments about [general] causation”
`based on a study or studies that find an association). The MDL court discussed this
`same issue in its analysis of the “analytical gaps” between that study and the experts’
`opinions. Accordingly, we consider this point in assessing the MDL court’s
`“analytical gaps” analysis.
`
`-7-
`
`

`

`
`
`A.
`
`As the proponent of the expert testimony in question, Plaintiffs have the
`burden to prove its admissibility by a preponderance of the evidence. Lauzon v.
`Senco Prods., Inc., 270 F.3d 681, 686 (8th Cir. 2001). Federal Rule of Evidence 702
`governs the admissibility of expert testimony, and under this rule the district court
`is “vested with a gatekeeping function, ensuring that ‘any and all scientific testimony
`or evidence admitted is not only relevant, but reliable.’” Union Pac. R.R. v. Progress
`Rail Servs. Corp., 778 F.3d 704, 709 (8th Cir. 2015) (quoting Daubert v. Merrell
`Dow Pharms., Inc., 509 U.S. 579, 589 (1993)). In exercising this gatekeeping
`function, the district court has “broad discretion,” and “on appeal we will not disturb
`a decision concerning the exclusion of expert testimony absent an abuse of that
`discretion.” Wagner v. Hesston Corp., 450 F.3d 756, 758 (8th Cir. 2006).
`
`That said, we have recognized that the “liberal thrust” of Rule 702 regarding
`the admissibility of expert testimony creates “an intriguing juxtaposition with our
`oft-repeated abuse-of-discretion standard of review.” Johnson v. Mead Johnson &
`Co., 754 F.3d 557, 562 (8th Cir. 2014). “While we adhere to this discretionary
`standard for review of the district court’s Rule 702 gatekeeping decision, cases are
`legion that, correctly, under Daubert, call for the liberal admission of expert
`testimony.” Id. (collecting authorities).
`
`
`Rule 702’s “screening requirement” has been “boiled down to a three-part
`test.” Id. at 561. First, the testimony must be useful to the finder of fact in deciding
`the ultimate issue of fact, meaning it must be relevant. See id. Second, the expert
`must be qualified to assist the finder of fact. Id. Third, the testimony must be
`reliable or trustworthy in an evidentiary sense. Id. At issue here is the third part of
`this test—whether Plaintiffs’ experts’ proposed testimony meets Rule 702’s
`reliability requirement. “The standard for judging the evidentiary reliability of
`expert evidence is ‘lower than the merits standard of correctness.’” Kuhn v. Wyeth,
`Inc., 686 F.3d 618, 625 (8th Cir. 2012) (quoting In re Paoli R.R. Yard PCB Litig.,
`35 F.3d 717, 744 (3d Cir. 1994)).
`
`-8-
`
`

`

`The reliability inquiry is a “flexible” one, with “[m]any factors” bearing on it.
`Daubert, 509 U.S. at 593-94. In Daubert, the Court articulated “four non-exclusive
`factors” relevant to this inquiry. Johnson, 754 F.3d at 562. These factors are
`(1) whether the expert’s theory or technique can be or has been tested, (2) whether
`the theory or technique has been subjected to peer review or publication, (3) the
`known or potential rate of error of the theory or technique, and (4) whether the
`technique or theory is generally accepted. See id.; Peitzmeier v. Hennessy Indus.,
`Inc., 97 F.3d 293, 297 (8th Cir. 1996). Factors recognized since Daubert include
`“whether the experts are proposing to testify about matters growing naturally and
`directly out of research they have conducted independent of the litigation, or whether
`they have developed their opinions expressly for purposes of testifying.” Daubert
`v. Merrell Dow Pharms., Inc., 43 F.3d 1311, 1317 (9th Cir. 1995).
`
`Additionally, while Daubert instructed that the focus of the reliability inquiry
`“must be solely on principles and methodology, not on the conclusions that they
`generate,” 509 U.S. at 595, the Supreme Court later clarified that “conclusions and
`methodology are not entirely distinct from one another,” Gen. Elec. Co. v. Joiner,
`522 U.S. 136, 146 (1997). Thus, “a district court’s focus on principles and
`methodology need not completely pretermit judicial consideration of an expert’s
`conclusions,” Kuhn, 686 F.3d at 625 (internal quotation marks omitted), and a
`district court may exclude expert testimony if it finds “that there is simply too great
`an analytical gap between the data and the opinion proffered,” Joiner, 522 U.S. at
`146. Or, to put it in the language we have frequently used both before and after
`Daubert and Joiner, a district court may exclude an expert’s opinion if it is “so
`fundamentally unsupported” by its factual basis “that it can offer no assistance to the
`jury.” E.g., Loudermill v. Dow Chem. Co., 863 F.2d 566, 570 (8th Cir. 1988); United
`States v. Finch, 630 F.3d 1057, 1062 (8th Cir. 2011).
`
`
`When a district court excludes an expert’s opinion for being fundamentally
`unsupported, yet another “intriguing juxtaposition” is evident in our case law. See
`Johnson, 754 F.3d at 562. On the one hand, we have recognized that we owe
`“significant deference” to the district court’s “determination that expert testimony is
`
`-9-
`
`

`

`excessively speculative,” and we “can reverse only if we are convinced that the
`District Court made a clear error of judgment on the basis of the record before it.”
`Grp. Health Plan, Inc. v. Philip Morris USA, Inc., 344 F.3d 753, 760 (8th Cir. 2003)
`(internal quotation marks omitted). On the other hand, we have stated numerous
`times that, “[a]s a general rule, the factual basis of an expert opinion goes to the
`credibility of the testimony, not the admissibility.” E.g., United States v. Coutentos,
`651 F.3d 809, 820 (8th Cir. 2011); see also Klingenberg v. Vulcan Ladder USA,
`LLC, 936 F.3d 824, 829-30 (8th Cir. 2019) (distinguishing cases where we affirmed
`the exclusion of experts’ opinions as too speculative because, in those cases, the
`experts’ opinions were “wholly speculative,” “connected to the facts by only the
`expert’s ipse dixit,” “patent speculation,” “pure conjecture,” and “vague theorizing
`based upon general principles”).
`
`Thus, excluding an expert’s opinion for being fundamentally unsupported is
`an exception to the general rule that “[g]aps in an expert witness’s . . . knowledge”
`go to weight, not admissibility. See Robinson v. GEICO Gen. Ins., 447 F.3d 1096,
`1100 (8th Cir. 2006); cf. Finch, 630 F.3d at 1062 (“Doubts regarding whether an
`expert’s testimony will be useful should generally be resolved in favor of
`admissibility.” (brackets omitted)). “Vigorous cross-examination, presentation of
`contrary evidence, and careful instruction on the burden of proof are the traditional
`and appropriate means” of addressing “shaky but admissible evidence.” Daubert,
`509 U.S. at 596.
`
`
`With these principles in mind, we analyze the reasons provided by the MDL
`court for excluding Plaintiffs’ general-causation medical experts and Dr. Elghobashi
`(whose CFD model and testimony formed part of the factual basis for the general-
`causation medical experts’ opinions).
`
`
`B.
`
`
`We first consider the MDL court’s determination that “too great an analytical
`
`gap” existed between “the literature” and Plaintiffs’ medical experts’ general-
`
`-10-
`
`

`

`causation opinions. This literature falls generally into two categories: (1) an
`epidemiological study reporting an association between forced-air warming and
`PJIs; and (2) studies and reports ostensibly supporting both of Plaintiffs’ mechanistic
`theories of causation. We consider this evidence to assess whether the MDL court
`“made a clear error of judgment on the basis of the record before it” in finding the
`experts’ opinions too speculative to be admitted. See Grp. Health Plan, 344 F.3d at
`760 (internal quotation marks omitted).
`
`
`
`
`1.
`
`All three medical experts relied on a 2011 observational epidemiological
`study as support for their conclusion that the Bair Hugger is capable of causing PJIs.
`See P.D. McGovern et al., Forced-Air Warming and Ultra-Clean Ventilation Do Not
`Mix, 93-B J. Bone & Joint Surgery 1537 (2011) (“McGovern 2011”). As an
`observational epidemiological study, McGovern 2011 explored whether forced-air
`warming was associated with an increased rate of PJIs by comparing a group of
`individuals warmed convectively to a group of individuals warmed conductively.
`See id. at 1537. See generally Reference Manual, supra, at 552, 555-56.
`
`
`Specifically, McGovern 2011 reviewed infection data from 1,437 hip- or
`knee-replacement surgeries performed at a particular hospital for a 2.5-year period.
`Id. at 1537, 1540. From July 2008 to March 2010, the patients were warmed with
`Bair Huggers; from March 2010 to June 2010, the hospital gradually transitioned to
`using conductive patient-warming devices; and from June 2010 to the end of the
`study, the patients were warmed solely with conductive patient-warming devices.
`Id. at 1540, 1543. The investigators found that patients warmed convectively were
`nearly four times more likely to contract a PJI than patients warmed conductively.
`Id. at 1541. The authors of McGovern 2011 acknowledged that the study did “not
`establish a causal basis” for this association. Id. at 1543. And they acknowledged
`
`-11-
`
`

`

`that their findings may have been “confounded”4 by “other infection control
`measures instituted by the hospital” during the study period (specifically identifying
`two such potentially confounding measures) and that they were “unable to consider
`all [patient-medical-history] factors” associated with PJIs, including a number of
`“important predictors for deep infection,” due to limited data in the records they
`reviewed. Id.
`
`The MDL court found that McGovern 2011 itself was sufficiently reliable to
`
`be admitted. But the MDL court faulted the experts’ reliance on it in ways that
`contributed to the analytical gap it found. For instance, the MDL court deemed it
`unreliable for the experts to draw an inference of causation from this study when the
`study disclaimed having proved causation. The MDL court also faulted how the
`experts handled the study’s limitations.
`
`As for the first point, we disagree that it is per se unreliable for an expert to
`
`draw an inference of causation from an epidemiological study that disclaimed
`proving causation. “[E]pidemiology cannot prove causation.” Reference Manual,
`supra, at 598. Instead, epidemiology enables experts to find associations, which by
`themselves do not entail causation. See id. at 552-53, 598. But an observational
`study such as McGovern 2011 “can be brought to bear” on the question of causation,
`id. at 217, and “can be very useful” to answering that question, id. at 221. Ultimately
`“causation is a judgment for epidemiologists and others interpreting the
`epidemiologic data.” Id. at 598; see also id. at 222 (“In the end, deciding whether
`associations are causal typically . . . rests on scientific judgment.”). Thus, it was not
`necessarily unreliable for the experts to rely on McGovern 2011 to draw an inference
`of causation just because the study itself recognized, consistent with these principles,
`that the association did not establish causation. So long as an expert does the work
`
`4“Confounding occurs when another causal factor (the confounder) confuses
`
`the relationship between the agent of interest and outcome of interest.” Reference
`Manual, supra, at 591. For instance, if those who drink alcohol are more likely to
`smoke than those who do not, then smoking may be a confounder in a study finding
`an association between drinking alcohol and emphysema. See id. at 592.
`
`-12-
`
`

`

`“to bridge the gap between association and causation,” a study disclaiming having
`proven causation may nevertheless support such a conclusion. See id. at 218.
`
`
`We recognize that there is language from Joiner that, when taken out of
`context, might appear to suggest otherwise. See 522 U.S. at 145 (“Given that [the
`authors of the study in question] were unwilling to say that PCB exposure had caused
`cancer among the workers they examined, their study did not support the experts’
`conclusion that Joiner’s exposure to PCB’s caused his cancer.”). But the context
`indicates that the problem with the experts’ opinions in that case was that they failed
`to bridge the gap left by the study in question. See id. at 145-46 (recounting
`numerous issues with the experts’ factual basis).
`
`As for the second issue, the MDL court rightly faulted the experts for how
`
`they handled McGovern 2011’s limitations. “Assessing whether an association is
`causal requires an understanding of the strengths and weaknesses of the study’s
`design and implementation,” and “the key questions” in evaluating epidemiological
`evidence “are the extent to which a study’s limitations compromise its findings and
`permit inferences about causation.” Reference Manual, supra, at 553. The experts
`did not adequately address McGovern 2011’s limitations. Neither Dr. Jarvis nor Dr.
`Stonnington mentioned the identified potential confounders or limitations in
`McGovern 2011 in their reports. Dr. Samet, on the other hand, did address in his
`report the two potential confounders identified by the authors in the study itself, and
`he meaningfully explained why in his view these variables did not confound the
`study’s findings. But Dr. Samet did not meaningfully address the other limitations
`identified by the McGovern 2011 authors except to say that confounding by other
`factors “seems unlikely” for ostensibly logical reasons.5
`
`
`5In the context of discussing this point, the MDL court also faulted Dr. Samet
`
`for departing “from his own description of reliable methodology.” Specifically, it
`noted that, with respect to the criterion of consistency that experts use in making
`judgments about causation, see Reference Manual, supra, at 600, Dr. Samet noted
`that that criterion “is generally applied as a consideration related to interpretation of
`findings of multiple observational studies and hence is not applicable to the single
`
`-13-
`
`

`

`However, McGovern 2011 was not the only basis on which the experts relied
`in forming their opinions. In addition to the epidemiological data from McGovern
`2011, the experts also relied on studies and reports ostensibly showing plausible
`mechanisms by which forced-air warming can cause PJIs. See id. at 599-600, 604
`(identifying the “biological plausibility” of a general-causation theory as one factor
`guiding epidemiologists in “making judgments about causation” and noting that
`“[w]hen biological plausibility exists, it lends credence to an inference of
`causality”). Thus, the experts’ failure to handle McGovern 2011’s limitations
`properly is not fatal to the admissibility of their opinions. See id. at 599-600 (listing
`“factors that guide epidemiologists in making judgments about causation”; noting
`that “there is no threshold number that must exist”; and including among these
`factors “[b]iological plausibility” along with “[c]onsideration of alternative
`explanations”). “[A]n inference of causation based on the totality of the evidence”
`may be reliable even if “no one line of evidence support[s] a reliable inference of
`causation” by itself. Milward v. Acuity Specialty Prods. Grp., Inc., 639 F.3d 11, 23
`(1st Cir. 2011); see also United States v. W.R. Grace, 504 F.3d 745, 765 (9th Cir.
`2007) (noting that whether an expert’s opinion testimony satisfies Rule 702
`“requires consideration of the overall sufficiency of the underlying facts and data”).
`Accordingly, we turn to the other evidence these experts considered.
`
`
`
`study by McGovern.” Because there was only one epidemiological study here, Dr.
`Samet instead “point[ed] to the consistency of the findings of studies addressing the
`effect of the Bair Hugger device on particle counts at the surgical site.” In context,
`we do not read this as Dr. Samet misapplying his own methodology but rather
`acknowledging that the consistency factor was not relevant in its conventional sense
`but nevertheless the consistency of the mechanistic studies supported an inference
`of causation from McGovern 2011’s finding of an association. As he explained in
`the next paragraph of his report following the language the MDL court quoted, the
`mechanistic evidence is consistent with McGovern 2011’s findings. The Reference
`Manual directs epidemiologists to consider whether the finding of an association is
`“consistent with other relevant knowledge,” supra, at 606, and we fail to see how
`mechanistic evidence would not count as “other relevant knowledge.”
`
`-14-
`
`

`

`
`
`2.
`
`The studies and reports ostensibly showing the “biological plausibility” of the
`medical experts’ general-causation opinions broadly fall into two categories. One
`set ostensibly supports Plaintiffs’ airflow-disruption theory, and the other ostensibly
`supports Plaintiffs’ dirty-machine theory.
`
`
`
`
`a.
`
`All three medical experts relied on Plaintiffs’ airflow-disruption theory as a
`plausible causal mechanism to support their general-causation opinions. Again,
`according to this theory, waste heat generated by the Bair Hugger creates convection
`currents that disrupt laminar airflow in operating rooms and transmit nonsterile air
`to the surgical site, causing PJIs. The experts relied on two general categories of
`evidence for this theory: (i) Dr. Elghobashi’s CFD model and (ii) published studies
`examining airflow patterns in operating rooms as well as the correlation between
`particles and bacteria. We consider each category in turn.
`
`
`
`
`i.
`
`To investigate whether forced-air warming “play[s] a role” in transporting
`squames (skin flakes capable of carrying bacteria that are present in operating
`rooms) to the surgical site, Dr. Elghobashi prepared a CFD model using large eddy
`simulation (a way to model fluid turbulence) to simulate the Bair Hugger’s effect on
`airflow and dispersion of squames in an ultra-clean-ventilation operating room. Dr.
`Samet and Dr. Jarvis relied on Dr. Elghobashi’s model. The parties agree, and the
`MDL court found, that the physics underlying Dr. Elghobashi’s model is reliable.
`Dr. Elghobashi eventually published his model with several coauthors in a peer-
`reviewed journal. See X. He et al., Effect of Heated-Air Blanket on the Dispersion
`of Squames in an Operating Room, 34 Int’l J. Numerical Methods Biomedical
`Eng’g, May 2018, at 1 (“He 2018”).
`
`
`-15-
`
`

`

`Dr. Elghobashi’s model replicated an orthopedic operating room, including
`details such as laminar airflow, an operating table, surgical drapes, a patient
`underneath the drapes prepared for knee surgery, four surgeons (two with hands
`extended over the patient, two with hands down), t