OPINION

O'SCANNLAIN, Circuit Judge:

We are called upon to decide whether, in August 2004, a manufacturer of electronic control devices, commonly referred to as "tasers," was under a duty to warn that repeated exposure to its products could lead to fatal levels of metabolic acidosis.

I

Shortly after 11:00 p.m. on August 29, 2004, a resident of Del Rey Oaks, California, called the police to report that someone who "look[ed] pretty disturbed" was "walking up and down the street yelling `Mario' and ... some other stuff." Del Rey Oaks Police Officer Russell Van Zanten responded to this noise complaint and found Michael Rosa ("Michael") still in the street. Believing that Michael was "either really high or crazy," Van Zanten approached him with considerable caution. Without fully exiting his patrol vehicle, Van Zanten identified himself as a police officer. Michael circled the vehicle — striking the hood with his hand as he passed — and began staggering down the street. Van Zanten followed in his vehicle, calling for assistance from surrounding law enforcement agencies.

Officer Jack Jeffrey Powell was the first of at least six additional officers to arrive. Due to Powell's angle of approach, Michael found himself between two patrol cars facing each other. At this point, he started "moving really awkwardly, kind of just like freaking out." The situation deteriorated rapidly as more officers arrived and as Michael attempted to flee, breaking at least one fence in the process.

The pursuit ended when Michael jumped over a three-foot fence, assumed a batter's stance, and began swinging a piece of two-by-four. Unable to convince Michael to comply with verbal commands and concerned for his safety, Officer Matthew Doza deployed his ADVANCED TASER M26 ECD ("M26") — a weapon known as an electronic control device that is manufactured by defendant-appellee TASER International, Inc. ("TASER"1) — in "probe mode."2 Michael then tumbled down an embankment, breaking the leads and ending the shock. Officer Doza followed, loading a new cartridge as he went. When Michael appeared to be reaching for the two-by-four again, Officer Doza redeployed his M26. Michael was apparently not incapacitated by this deployment. But, noting some effect on him, Doza depressed the trigger an additional six or seven times. Seeing that Michael continued to struggle, newly arrived Officer Nicholas Borges deployed his own M26. He cycled his M26 three times before Michael finally hit the ground.

Nevertheless, Michael continued to resist as yet another officer, Katie Reyes, attempted to place him in handcuffs. Including Officer Reyes, it took the efforts of six officers to subdue him. And in order finally to take him into custody, Doza once again had to apply his M26, this time in "drive-stun mode" to Michael's upper leg.3 Until the handcuffs were in place, Michael continued to struggle and did not appear to be in medical distress.

After officers had Michael in restraints, they rolled him onto his side. At this point, Michael slumped, his lips blue, his breathing erratic. He quickly stopped breathing entirely. Officers were unable to find a pulse and immediately began resuscitation efforts. Michael was transported to the hospital, where resuscitation efforts continued. But Michael's heart went into atrial arrhythmia (a form of irregular heartbeat), tachycardia (accelerated heartbeat), and finally asystole (cardiac arrest). He was pronounced dead at about 12:30 a.m. on August 30, 2004.

Dr. John Hain performed the autopsy. Discovering high levels of methamphetamine in Michael's blood, Dr. Hain concluded that his cause of death was "ventricular arrhythmia ... due to methamphetamine intoxication." He listed "Taser application and arrest by police" as contributing conditions. Michael's death was subsequently linked to metabolic acidosis, a condition under which lactic acid — a byproduct of physical exertion — accumulates more quickly than the body can dispose of it, causing the pH in the body to decrease. The condition makes sudden cardiac arrest more likely.

Plaintiffs-appellants, Michael's parents, Evelyn and Robert, and his daughter Holly (collectively "Rosas") brought this lawsuit against TASER, as manufacturer of the M26, asserting that Michael died because it had provided an inadequate warning of the dangers of the product to the officers who used it.4 They pursued both strict liability and negligence theories under California law based upon this failure to warn. At the times in question, TASER provided warnings that read in relevant part:

The Rosas claimed that TASER also should have warned that repeated exposure to the M26 carried its own risks, particularly the risk that it can cause fatal levels of metabolic acidosis.

After the conclusion of discovery, the district court awarded summary judgment to the defendant. The court concluded that the Rosas had not established a triable issue of fact that the risk of metabolic acidosis was known or knowable when the M26 at issue was distributed in December 2003 (or even when Michael died the following August). Rosa v. City of Seaside, 675 F.Supp.2d 1006, 1013-14 (N.D.Cal. 2009) (order). It reasoned that the scientific research cited by the Rosas did not address TASER's products, was not publicly available, or represented nothing more than hypotheses, unproven by scientific methodology. Id. As a result, it concluded that the manufacturer was not liable under strict liability. Id. at 1014. The district court concluded based on similar flaws of proof that the Rosas had not established a triable issue of fact that TASER should have known of the risk. Thus, it was not liable for negligence. Id. at 1015. The Rosas timely appealed.

II

The Rosas first argue that the district court construed TASER's duty to warn too narrowly under California's strict liability rules.5 They assert that TASER had a broader duty to test the risks of its products, as well as to integrate into its warnings information gleaned from isolated reports of potential side effects.

A

California law places a duty on manufacturers to warn of a "particular risk" if it is "known or knowable in light of the generally recognized and prevailing best scientific and medical knowledge available at the time of manufacture and distribution." Conte v. Wyeth, Inc., 168 Cal.App.4th 89, 85 Cal.Rptr.3d 299, 310 (2008) (emphasis added). Thus, this case turns on what was "knowable" by a manufacturer of electronic control devices in December 2003. The Rosas argue essentially that any risk that was discoverable through modern technology, no matter how unsubstantiated, was knowable by TASER. We do not interpret the standard so broadly.

Though the California courts have never announced a comprehensive standard of when a particular risk is "knowable," a few key considerations are clear. "[A] manufacturer is held to the knowledge and skill of an expert in the field; it is obliged to keep abreast of any scientific discoveries and is presumed to know the results of all such advances." Carlin v. Superior Court, 13 Cal.4th 1104, 56 Cal.Rptr.2d 162, 920 P.2d 1347, 1351 n. 3 (1996). A manufacturer cannot defeat liability because it did not review the relevant scientific literature.

But a manufacturer is not under a duty to warn of "every report of a possible risk, no matter how speculative, conjectural, or tentative," because "inundat[ing the public] indiscriminately with notice of any and every hint of danger" would "inevitably dilut[e] the force of any specific warning given." Finn v. G.D. Searle & Co., 35 Cal.3d 691, 200 Cal.Rptr. 870, 677 P.2d 1147, 1153 (1984); see also Brown v. Superior Ct., 44 Cal.3d 1049, 245 Cal.Rptr. 412, 751 P.2d 470, 480-81 (1988) (noting that strict liability is not designed to turn manufacturers into insurers of their product). For example, in Finn v. G.D. Searle & Co., doctors prescribed diodoquin to a child diagnosed in 1969 with acrodermatitis enteropathica after experiencing a severe rash and diarrhea. In 1971, the child developed vision problems caused by optic nerve atrophy. His parents sued the manufacturer of diodoquin for a failure to warn. To show that the risk that diodoquin could cause optic nerve atrophy was knowable at the time of distribution, the Finns presented a 1966 article published in the British medical journal, The Lancet, reporting another case of a child developing optic nerve atrophy after long term use of diodoquin for the same condition. 200 Cal.Rptr. 870, 677 P.2d at 1149-50. They also presented a 1971 report implicating a drug in the same chemical family in causing optic nerve atrophy. Id. In that case, the defendants were not held liable for failure to warn, and the California Supreme Court clarified that "[k]nowledge of a potential side effect which is based on a single isolated report of a possible link between a [product] and an injury may not require a warning." Id. at 1153. That is, such reports of isolated or speculative injuries do not constitute generally accepted medical knowledge.

B

1

To establish that the risk that TASER's products could cause fatal levels of metabolic acidosis was knowable by December 2003, the Rosas rely primarily on four peerreviewed articles. We conclude that these articles do not present a triable issue of fact that the risk was more than purely speculative.

First, the Rosas point to a 1966 study of the impact of acidosis on the risk of ventricular fibrillation (a form of cardiac arrhythmia). See Paul H. Gerst, et al., Increased Susceptibility of the Heart to Ventricular Fibrillation During Metabolic Acidosis, 19 CIRCULATION RESEARCH 63 (1966). This study demonstrates nothing more than that the risk of ventricular fibrillation increases as blood pH decreases. Id. at 63. But the study makes no attempt to link either its findings or acidosis to the use of electronic control devices. Furthermore, the authors themselves warn that "it may not be justifiable to extrapolate conclusions derived from an experimental study such as this to human physiology." Id. at 68. Given these limitations, this article did not place TASER on notice that its products could cause cardiac problems in humans.

Second, the Rosas point to a 1999 case study examining the link between metabolic acidosis and deaths that occur while in police custody. See John L. Hick, et al., Metabolic Acidosis in Restraint-Associated Cardiac Arrest: A Case Series, 6 ACAD. EMERGENCY MED. 239 (1999). This study was one of a number performed to determine the causes of a phenomenon known as "sudden in-custody death syndrome." Id. at 243. As the study pointed out, "[t]he phenomenon of sudden death in restraints has long been recognized," but "[t]he physiologic derangements that lead to death are still poorly understood." Id. at 241. Proposed hypotheses included "autonomic reflexes, arrhythmias, or restraint stress" as well as positional asphyxia, "cocaine ingestion[,] and significant exertion." Id. This study sought to examine the link between acidosis and these deaths. The authors' main thesis was that prolonged struggle with police increases the risk of death by causing acidosis; their only conclusion was that police should eschew the "hobble restraint position" in favor of placing suspects on their sides. Id. at 242. Not only was the study incomplete because "[l]actate levels," an indicator of acidosis, "were obtained in only one" of the five cases studied, but none of the cases involved electronic control devices. Id. at 240. As the M26 was designed specifically to prevent the sort of prolonged struggle that the article posed as the main risk for acidosis, this article did not put TASER on notice that its product might cause similar injuries.

Third, the Rosas point to an article published in The Lancet in 2001, which they claim provides the missing link between electronic control devices and metabolic acidosis and thus to cardiac arrhythmia. See Raymond Fish & Leslie A. Geddes, Effects of Stun Guns and Tasers, 358 THE LANCET 687 (2001). Finding the answers posed by previous attempts to explain sudden in-custody death syndrome unsatisfying, Drs. Fish and Geddes hypothesized that electronic control devices may contribute to the condition by "affect[ing] acid-base balance" of the individuals exposed to them. Id. at 688. However, they made no attempt to test the hypothesis and noted that there was "no adequate information" to link these deaths to exposure to TASER's products. This sort of hypothetical side effect is insufficient to require a warning under California law. Finn, 200 Cal.Rptr. 870, 677 P.2d at 1153.

Fourth, the Rosas point to a study that was performed on behalf of the Department of Defense in 1999, but that did not become publicly available until after Michael's death. JOHN M. KENNY, ET AL., HUMAN EFFECTS ADVISORY PANEL, REPORT OF FINDINGS STICKY SHOCKER ASSESSMENT (1999). Assuming that a document that is not publicly available can constitute generally accepted medical knowledge — which we doubt — this study suffers the same problem as the article from The Lancet. It merely states that "deaths following Taser[] use may be due to acidosis." Id. at 31. It does not purport to establish that causal link and explicitly limits the reach of its findings due to its small data set. Id. at 32.

Because these last two articles provided the only link between electronic control devices and severe levels of acidosis, even taken with the Gerst and Hick studies, they do not establish a triable issue of fact that the risk of metabolic acidosis was knowable at the time of distribution.

2

The Rosas attempt to avoid this outcome by pointing to a warning TASER issued in 2009 that explicitly discusses the risk of metabolic acidosis.6 In examining whether summary judgment is appropriate, we "consider only alleged facts that would be admissible in evidence." Filco v. Amana Refrigeration, Inc., 709 F.2d 1257, 1260 (9th Cir.1983) (citing FED. R. CIV. P. 56(e)). Because the fact of this 2009 warning is not admissible to establish what was knowable in December 2003, it cannot aid the Rosas in avoiding summary judgment. FED. R. EVID. 407; see also Gauthier v. AMF, Inc., 788 F.2d 634, 636 (9th Cir. 1986) (applying Rule 407 in a products liability case under Montana law); Flaminio v. Honda Motor Co., 733 F.2d 463, 471 (7th Cir.1984) (explaining that Rule 407 was sufficiently procedural in nature to apply in diversity cases under Erie R.R. Co. v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938)).7

III

The Rosas next assert that the district court erred in awarding summary judgment to TASER on their negligence claim. Though they do not precisely lay out how their negligence claim differs from their strict liability claim, they point to repeated statements by the California courts that the duties to warn under strict liability and negligence theories are not coterminous. See, e.g., Valentine v. Baxter Healthcare Corp., 68 Cal.App.4th 1467, 81 Cal.Rptr.2d 252, 263 (1999). While this is true, there is also considerable overlap between the two torts. Indeed, generally a "manufacturer's duty, per strict liability... to warn of potential risks and side effects envelopes a broader set of risk factors than the duty, per negligence ... to warn of facts which make the product `likely to be dangerous' for its intended use." Id. (emphasis in original). In this case, we see three potential circumstances when negligence might provide a broader duty to warn than does strict liability, but we conclude that the Rosas have not established a triable issue of fact on any of them.

First, under certain circumstances, California's negligence law may impose on a manufacturer a duty to warn individuals who, while not users of its products, could foreseeably rely on its warnings. For example, when the user of a generic pharmaceutical sues the manufacturer of the brand name medication for the warning included in the Physician's Desk Reference, the user cannot recover under strict liability because he or she was not injured by the manufacturer's own products. Conte, 85 Cal.Rptr.3d at 309-10, 316. However, because the brand name manufacturers are responsible for disseminating the information in the Physician's Desk Reference, which others would foreseeably rely upon, they may be held liable under negligence. Id. at 317-18. Such a duty thus arises only in narrow circumstances that are not present here.

Second, though California law measures the strict liability duty to warn from the time a product was distributed, a manufacturer may be liable under negligence for failure to warn of a risk that was subsequently discovered. Oxford v. Foster Wheeler LLC, 177 Cal.App.4th 700, 99 Cal.Rptr.3d 418, 432-33 (2009) (discussing Hernandez v. Badger Constr. Equip. Co., 28 Cal.App.4th 1791, 34 Cal.Rptr.2d 732 (1994)). However, the record indicates only one development between the date that this M26 was distributed and Michael's death: TASER became aware of the study regarding the "Sticky Shocker" done for the Department of Defense.8 As discussed above, that study merely states an untested hypothesis. This does not create a triable issue of fact that a reasonable manufacturer would have sent a supplemental warning based on this information.

Third, under California law, TASER may be liable in negligence for failure to undertake sufficient testing before distribution. Cf. Valentine, 81 Cal. Rptr.2d at 264. But the Rosas have put forth no evidence creating an issue of fact regarding whether it conducted reasonable testing. Before Michael's death, the perceived cardiac risk associated with the device was immediate ventricular fibrillation, and TASER expended considerable resources testing its products for that risk.9 This does not establish a triable issue of fact that TASER failed to exercise due care when this new risk was brought to its attention.

Because California law would not extend negligence liability to a manufacturer in this circumstance, the district court properly awarded summary judgment.

IV

For the foregoing reasons, we conclude that the district court properly awarded summary judgment in favor of TASER because the risk of lactic acidosis was not knowable in 2003. Thus, we do not reach TASER's alternative arguments.

AFFIRMED.