Appeal from the United States District Court for the Northern District of Illinois, Eastern Division. Nos. 1:22-cv-00071 & 1:22-cv-00232 — Rebecca R. Pallmeyer , Judge .

ARGUED MAY 20, 2026 — DECIDED JULY 24, 2026

Before BRENNAN, Chief Judge , and SCUDDER and JACKSONAKIWUMI, Circuit Judges .

BRENNAN, Chief Judge . In 2014 RaiLee Mar was born 12 weeks prematurely. Initially, she was fed her mother’s milk. But over a week later, her mother could no longer produce usable milk, and the hospital had no donor milk. So, doctors Page 2fed RaiLee Similac Special Care 24-a cow's-milk-based infant formula produced by Abbott . A day later, RaiLee was diagnosed with a gastrointestinal disease, necrotizing enterocolitis (“NEC”). The next day, tragically, she died.

RaiLee's mother Ericka Mar sued Abbott , alleging the company failed to warn about the dangers associated with its formula, including that it could cause NEC. Her case joined hundreds of others in multidistrict litigation and was selected as a bellwether. After discovery, the district court granted Abbott's motion for summary judgment , and Mar appeals. We affirm because she has not shown that the warning she believes Abbott should have used would have prevented her daughter's death.

NEC is a serious “gastrointestinal disease that commonly affects preterm infants and is a major cause of morbidity and mortality.' Dhirendra Singh et al., Necrotizing Enterocolitis: Bench to Bedside Approaches and Advancing Our Understanding of Disease Pathogenesis , FRONTIERS IN PEDIATRICS, Jan. 11, 2023, at 1, 1. Symptoms include swelling of the abdomen, vomiting, and gas within the walls of the bowel, all of which can cause death. DEP'T OF HEALTH AND HUM. SERVS., NECROTIZING ENTEROCOLITIS (NEC) IN PRETERM INFANTS, 1 (2024) ('HHS NEC”). Though newborns of any gestational age can develop NEC, scientists believe that preterm infants are especially at risk because of their underdeveloped immune systems. Singh, supra at 3.

“One of the ways to decrease NEC incidence is to provide maternal breast milk to infants.” Id. at 2. Studies in published guidelines from the American Academy of Pediatrics ('the Page 3Academy”) show a 58% decrease in NEC when breast milk was fed to preterm infants. See the Academy, Breastfeeding and the Use of Human Milk , 129 PEDIATRICS e827, e829 (2012). Another study found a 77% reduction in NEC in preterm infants fed exclusively breast milk. Id. According to Abbott's evidence, neonatologists have long known that breast milk confers unique nutritional benefits. Still, the link between NEC and breast milk is not fully understood.

Cow’s-milk-based formula appears to not have the same NEC-reducing traits as breast milk. For example, formula-fed babies were “6-10 times” more likely to develop NEC “than in those fed breast milk alone.” A. Lucas & T.J. Cole, Breast Milk and Neonatal Necrotizing Enterocolitis , 336 MED. SCI. 1519, 1519 (1990). There are two ways to interpret this link. Either formula causes NEC, or formula is less effective at preventing the disease than breast milk. Still, “there are substantial gaps in knowledge” on the question of causation and nutrition. See HHS NEC, supra , at 5.

Current scientific evidence recommends breast milk from the baby’s mother a s the first choice to feed preterm infants. The next best alternative is donor breast milk. Last, doctors turn to cow's-milk-based formulas. The Abbott formula challenged here comes with the warning, “USE AS DIRECTED BY A DOCTOR.”

On New Year’s Day 2014, RaiLee Mar was born at a hospital in Summersville, West Virginia. She arrived 12 weeks prematurely , putting her at high risk of developing NEC. She was airlifted to the neonatal intensive care unit at Charleston Area Medical Center. For the first four days, doctors fed RaiLee intravenously. On the fifth day, she was fed through a stomach tube. Over a week later, Mar could not produce any Page 4more usable breast milk; the small amount that she did was tainted with blood from over-pumping.

The hospital turned to the Abbott formula at issue because it had no donor milk. Another patient at the hospital who had recently given birth offered to donate some of her breast milk to RaiLee. But hospital policy forbade sharing milk because diseases can be transmitted without first treating the milk. So, RaiLee’s medical team changed her diet to 50% milk from Mar and 50% Abbott's formula. She was fed this mixed diet three times between January 14 and 15. On January 15, she was diagnosed with NEC. She passed away the next day.

About eight years later, in January 2022, Mar sued Abbott, alleging various theories of liability: strict liability design defect; negligence; and failure to warn. Her lawsuit joined hundreds of like cases against Abbott as part of a multi -district litigation proceeding. The parties selected Mar’s case as one of four bellwether cases to proceed to trial.

Abbott moved for summary judgment, which the district court granted. The court first concluded that Mar had not shown there was an alternative, feasible design for Abbott’s formula. Second, the court held that Mar had not produced sufficient evidence for its failure-to-warn theory. To prove this theory, Mar had to show that a different warning would have prevented RaiLee’s death. Her proposed warning would not have made a difference , the court reasoned, because the cow’s milk formula was the only available food source. Moreover, concluding that Mar’s alternative warning would have prompted the hospital to create a time- and resource-intensive donor milk program was speculative.

Page 5Mar moved to reconsider under Federal Rule of Civil Procedure 59(e), asking to present two new witnesses to support her failure-to-warn theory. Because both witnesses were available to Mar during discovery, the district court ruled she could not now offer them via a Rule 59(e) motion.

Mar appeals the failure-to-warn and Rule 59 rulings.

We review the district court’s grant of summary judgment de novo, construing the facts in the light most favorable to Mar and drawing all inferences in her favor. Lewis v. Ind. Dep’t of Transp. , 173 F.4th 876, 882 (7th Cir. 2026). But not 'every conceivable inference” is drawn in the non-movant’s favor. FKFJ, Inc. v. Village of Worth , 11 F.4th 574, 585 (7th Cir. 2021). Only those that are “reasonable” will be, and inferences “supported by only speculation or conjecture” are not reasonable. Osborn v. JAB Mgmt. Servs., Inc. , 126 F.4th 1250, 1258 (7th Cir. 2025).

West Virginia holds manufacturers liable for not warning consumers about their products' dangers.¹ Morningstar v. Black and Decker Mfg. Co. , 253 S.E.2d 666, 682 (W. Va. 1979). Liability for failure to warn, whether based on strict liability or negligence, requires a plaintiff to prove two elements. First, he must show “it was reasonably foreseeable to the manufacturer that the product would be unreasonably dangerous if distributed without a warning.” Church v. Wesson , 385 S.E.2d 393, 396 (W. Va. 1989). Second, the manufacturer’s failure to warn of the dangers must have been the cause in fact of the injuries. Tracy v. Cottrell ex rel. , 524 S.E.2d 879, 890 n.9 (W. Va.

Page 61999). Said otherwise, another warning must have “made a difference.' Id.

The warning on Abbott's formula stated, 'USE AS DIRECTED BY A DOCTOR.” Mar’s regulatory and labeling expert, Dr. Scheer, offered what he opined was a comparatively more effective and adequate warning: 'human milk has a lower risk of NEC than formula.” This label would have prevented RaiLee's death, Mar argues. Abbott disagrees; after all, the hospital had to feed RaiLee something, and there was nothing else on hand—the hospital had no donor milk, and Mar could not produce any more usable breast milk. So, the treating physicians had no other option but Abbott’s formula .

Mar replies that her alternative warning would have made a difference in three ways. And any reasonable counterfactual inferences, she reminds us, are “to be believed” and left for the jury to credit. Groh v. Ramirez , 540 U.S. 551, 562 (2004).

First, Mar claims that had Abbott used her alternative warning, the hospital would have made efforts to implement a donor-milk program. Donor milk then could have been fed to RaiLee. To support this counterfactual, she looks to the deposition testimony of Dr. Maxwell, RaiLee’s treating physician. She focuses on one exchange. Dr. Maxwell was asked, “[i]f you knew back in 2014 that formula was the primary risk factor or significantly increased the risk of NEC, would you have made efforts to get donor milk at that time?' He responded, in part, “we would have tried, yes, to get donor milk maybe quicker than we did.” This concession, Mar argues, creates a reasonable inference that her alternative warning would have spurred the creation of a donor-milk program.

Page 7Mar overreads Dr. Maxwell's testimony. Examining the full transcript, he believed that creating a donor-milk program would have been an immense challenge. Doing so, he explained, would be 'very expensive' and take years of effort. Indeed, it took five years before the hospital implemented a donor milk program, well after RaiLee's death; it was not something that happened “overnight.” Further, Dr. Maxwell was aware that the Academy recommends hospitals carry only breast milk, advice he believed “in the Appalachian area is probably impossible.” That is, in part, because the record shows several hurdles to clear before establishing a donormilk program: the type of containers, labeling techniques and transportation, personnel and training, storage, and thawing and warming methods.

Another hurdle to such a program is that approval is required from more than just the doctors. As Dr. Maxwell explained, the hospital board and foundation also had to agree to the program. So even if all the treating physicians had read the alternative warning label, without board approval, the record indicates that the donor milk program still may not have existed in time to save RaiLee.

Moreover, at the time RaiLee’s treating physicians knew of the connection between NEC and formula. Since 2011, Dr. Maxwell, a member of the Academy, stated he was “aware that formula feeding was associated with higher rates' of NEC. He admits he “must have read” the 2012 Statement on Breastfeeding and the Use of Human Milk. Dr. Shah, another of RaiLee’s treating physicians and also an Academy member, regularly read literature about the benefits of human milk and reducing NEC risks. So RaiLee’s doctors likely knew of the relationship between NEC and formula, and yet there was no Page 8donor milk program. Therefore, the proffered alternative warning, which links NEC and formula, would have made no difference.

To be sure, Dr. Maxwell did acknowledge that the hospital could have created a donor-milk program. But that single statement is a small fraction of his testimony on this topic. Given that Dr. Maxwell already knew of the association between NEC and formula, and his explanation of the great effort and expense necessary to obtain donor milk, it is not reasonable to infer that Mar’s alternative warning would have catalyzed the creation of such a program. Indeed, Dr. Maxwell stands by his treatment decisions for RaiLee.

Mar's inference, then, is speculative. And we 'will not draw inferences that are supported only by speculation.' In re Greenpoint Tactical Income Fund LLC , 168 F.4th 1002, 1008 (7th Cir. 2026). In the end, the district court correctly observed that Mar “overstates the strength” of Dr. Maxwell’s testimony.

Mar’s second argument focuses on what happened at the hospital. She was struggling to produce breast milk, and the milk she did pump contained blood, which she was told to discard. Another patient offered to share her breast milk . But sharing was against hospital policy because untested, unpasteurized milk can transmit diseases to the baby. Even so, Mar claims that had the hospital seen her alternative warning, physicians would have selected the unpasteurized milk instead of Abbott's formula.

I rrespective of Abbott’s warnings, the hospital prohibited the sharing of untested, unpasteurized breast milk. Dr. Maxwell explained why: “we don’t know the medical history of the person that’s providing the milk. For all that we know she Page 9might have hepatitis or HIV or something else.” As well, a nurse practitioner who testified acknowledged that the hospital did not want to “assum[e] the risk of the donor mother potentially transmitting disease ,” which is why hospital policy was to not share breast milk. So, as the district court correctly observed, Mar cannot point to evidence that the hospital would have changed its policy in light of a different warning on Abbott’s formula . Concluding otherwise requires conjecture.

Third, Mar claims that had Abbott used her alternative warning, the jury could have inferred that she would have declined formula and refused to discard her blood-tainted milk. This argument falls short, though, because Mar did not read Abbott’s original warning. West Virginia law requires a plaintiff first read the existing warning in failure-to-warn claims. See, e.g. , Howard v. Eaton Corp. , No. 16–0055, 2016 WL 6651592, at *4 (W. Va. Nov. 10, 2016); Meade v. Parsley , No. 2:09-cv00388, 2010 WL 4909435, at *9–10 (S.D. W. Va. Nov. 24, 2010); Shanklin v. Allis-Chalmers Mfg. , 383 F.2d 819, 823–24 (4th Cir. 1967) (West Virginia law). We do not know how Mar would react to reading the allegedly improper warning label, so we cannot determine how she would respond to reading a different one.

To Mar, the inferences required by all three of her arguments are reasonable. She offers Cloutier v. GoJet Airlines, LLC , 996 F.3d 426 (7th Cir. 2021), for the proposition that counterfactual inferences are for a jury to decide. The inferences the jury was permitted to weigh in that case, she contends, are the same as the inferences here.

In Cloutier , the plaintiff was diagnosed with diabetes and prescribed medication. Id. at 430. He missed the statutory Page 10deadline for his return-to-work clearance because his employer, GoJet, failed to provide him with the required notices. Id. at 440-41. But, he argued, 'had he been given the notices that GoJet was required to give him,' he would not have missed the deadline. Id. at 440 ( citation modified ). That is speculative, GoJet countered, because the plaintiff could not prove what might happen in an alternative world. Id. at 441.

We disagreed, emphasizing that counterfactual inferences are permissible only if sufficiently grounded in the record. Id. And importantly, GoJet offered no evidence to doubt those inferences. So, with evidence from the plaintiff, a jury could have drawn reasonable inferences in his favor. “GoJet argues [plaintiff] had no control over his doctors … [but] [plaintiff] testified that he could have received his approval in time if he had pushed his doctors. Absent concrete evidence to the contrary , a reasonable jury could have decided to credit that testimony.” Id. at 442 (emphasis added). And “GoJet argues that [plaintiff] also merely speculated about the FAA approving him sooner than September 4, 2014 … however, GoJet does not point to any evidence of its own undermining [plaintiff's] testimony.” Id.

So, Mar correctly reads Cloutier that counterfactuals can be submitted to the jury. But not all qualify. Only those with supporting record evidence and in which the defendant did “not point to any evidence of its own undermining the [plaintiff's] testimony' may reach the jury. Id. Cloutier does not help Mar because Abbott provide s strong evidence in support of its position.

Of course, we credit all of Mar’s evidence and acknowledge her arguments. And summary judgment is not a replacement for trials. Waldridge v. Am. Hoechst Corp. , 24 F.3dPage 11 918, 920 (7th Cir. 1994). Even so, the inferences she asks this court to draw from Dr. Maxwell’s statements and Mar’s alternative warning are too speculative to defeat summary judgment. Cf. Waukegan Potawatomi Casino, LLC v. City of Waukegan , 128 F.4th 871, 878 (7th Cir. 2025) (“string of inferences” cannot defeat summary judgment).

Abbott requests that we take one further step and rule on a broader topic of causation. The company asks us to conclude that Mar has not produced sufficient evidence to find that Abbott's formula was a 'but for' cause of NEC. As the district court observed, the answer to that question may turn on how RaiLee's feeding schedule is measured. We need not reach this weighty issue. For this appeal, it is sufficient to conclude that on this evidence, Mar’s suggested alternative label would not have prevented RaiLee’s death.

To Mar’s credit, she tried to bolster the inferences from her evidence with further proof . She filed a Rule 59(e) motion for reconsideration with the district court, referencing two pieces of 'newly discovered evidence.' First, RaiLee's father Anthony Mar explained that had he known about the risk of NEC from Abbott's formula, he would have pursued different options to feed RaiLee, such as through a donor breast milk system or taken her to a different hospital . Second, Dr. BarYam, a physician at a Boston-based hospital, stated that donor milk could be sent from her own milk bank to RaiLee’s hospital within 24 hours. The district court denied Mar’s motion, which we review for abuse of discretion. Reilly v. Will Cnty. Sheriff's Off. , 142 F.4th 924, 929 (7th Cir. 2025).

Page 12Rule 59(e) states, “[a] motion to alter or amend a judgment must be filed no later than 28 days after the entry of the judgment.” FED. R. CIV. P. 59(e). The movant must “present either newly discovered evidence or establish a manifest error of law or fact.” Oto v. Metro. Life Ins. , 224 F.3d 601, 606 (7th Cir. 2000). Mar argues the former—that her evidence is newly discovered.

But “[a] party may not use a motion for reconsideration to introduce new evidence that could have been presented earlier.” Id. Mar knew about her two witnesses at the time of summary judgment. RaiLee's father, known to Mar, could have been subpoenaed or ordered to testify. And Mar knew of Dr. Bar-Yam's proposed testimony before the summary judgment hearing. These witnesses, then, are not “newly discovered” evidence, and the district court did not abuse its discretion in denying Mar's motion. But even if the witnesses had been 'newly discovered,' the district court concluded that their testimony would not have moved the needle for Mar, a decision with which we have no quarrel.

* * *

This is a tragic case by any measure. But Mar has not shown that her alternative warning would have made a difference. And her motion to reconsider was properly denied.

AFFIRMED