On September 12, 2005, a 15-year-old athlete had surgery on her right knee. Afterward, a pain pump infused the local anesthetic bupivacaine directly into the joint for at least two days. Her knee soon became stiff and painful. Later x-rays showed that the cartilage in the knee had been destroyed.
She and a second young athlete sued the pump's manufacturer. Both alleged that the device caused chondrolysis, a permanent loss of joint cartilage. To support that link, they retained two epidemiologists. The manufacturer moved to exclude them. On January 24, 2011, a federal court in Ohio denied the motions.
Epidemiology is the study of disease patterns in human populations. Its experts ask whether people exposed to a substance develop a disease more often than people who were not exposed. Courts hear from them because that question sits at the center of general causation. Before a jury decides whether a product injured one person, it must decide whether the product can cause the injury at all.
This article reviews five rulings in which courts admitted epidemiological causation testimony over a Daubert or Rule 702 challenge. In one, the court admitted the opinion with a limit on its scope.
Where Epidemiology Experts Appear in Litigation
The five rulings come from drug, device, and environmental exposure cases. Two were multidistrict litigations over prescription drugs. In re Taxotere involved a chemotherapy drug, and In re Tasigna involved a leukemia drug. Hamilton v. Breg involved a medical device. Thompson v. Kinder Morgan Altamont alleged contamination from a neighboring industrial facility. Giles v. Wyeth alleged that an antidepressant led to a suicide.
In every case, the epidemiology testimony addressed general causation: whether a substance can cause a condition in the general population. In Giles, Plaintiff's expert also addressed specific causation, whether the drug caused one person's death.
The method followed a familiar two-step structure. First, the expert looked for an association between exposure and disease. The usual measures are relative risk and the odds ratio. A ratio of 1.0 means no association, and higher values mean a stronger one. Second, the expert asked whether the association was causal. In Taxotere, Tasigna, and Hamilton, the experts applied the Bradford Hill criteria to that question. Those criteria include strength of association, temporality, dose-response, consistency, and biological plausibility.
The challenges followed patterns too. Defendants argued that the underlying studies lacked statistically significant results. They argued that experts relied on studies of the wrong outcome, the wrong drug class, or the wrong type of exposure. They attacked the quality of the data, including unblinded trials and adverse event reports. One argued that the expert had to identify a threshold dose. Another pointed to a large federal analysis that found no effect.
Five Rulings That Admitted Epidemiology Testimony
In re Taxotere: Adverse Event Data and the Dose Question
In re Taxotere (Docetaxel) Eye Injury Products Liability Litigation, U.S. District Court for the Eastern District of Louisiana, MDL No. 3023, ruling dated April 2, 2026.
Plaintiffs in this MDL received the chemotherapy drug Taxotere, or docetaxel, mainly for breast cancer. They alleged that the drug narrowed their tear drainage ducts, a condition called stenosis. The alleged result was excessive tearing, known as epiphora. By the time of the ruling, Sanofi was the only remaining Defendant.
Plaintiffs' expert was a tenured professor and Vice Chair of the Department of Epidemiology at the University of Alabama at Birmingham. He held an M.S. from Harvard University and a Ph.D. from UAB. He had authored or co-authored more than 850 peer-reviewed manuscripts, many on eye disease, including work on docetaxel and epiphora. Sanofi did not challenge his qualifications.
He opined that the research supported a causal relationship between docetaxel and stenosis of the tear drainage system. His central study was his own 2023 analysis of the FDA Adverse Event Reporting System. It found a proportional reporting ratio of 19.54 for docetaxel and dacryostenosis, with a p-value below .001. He then applied the Bradford Hill criteria to the wider literature.
Sanofi raised four objections. It argued that a disproportionality analysis cannot establish an association, and that he needed a significant result for each dosing schedule. It argued that he had to identify a threshold dose. It also faulted his treatment of strength and background risk.
The Court rejected each one. It held that a reporting ratio can provide evidence of an association. The expert had acknowledged the method's limits and did not treat his study as definitive proof. On adverse event data, the Court wrote: "Although there are limitations with FAERS data, this does not warrant exclusion." On threshold dose, it relied on the Fifth Circuit's decision in Ruffin v. BP, which rejected that requirement for general causation.
On dose-response, the Court pointed to studies showing higher cumulative doses among patients with stenosis. Several of those studies were authored by Sanofi's own expert. On background risk, Plaintiffs' expert testified: "In this case, I didn't need to make that determination. The individual studies that I included would have made that determination." The Court accepted that the studies' comparison groups accounted for the baseline rate.
The Court denied the motion to exclude. It also admitted the general causation opinions of Plaintiffs' toxicologist. Sanofi's motion for summary judgment rested solely on the claim that Plaintiffs lacked admissible general causation testimony. Having found that testimony admissible, the Court denied summary judgment.
In re Tasigna: A Meta-Analysis of Randomized Trials
In re Tasigna (Nilotinib) Products Liability Litigation, U.S. District Court for the Middle District of Florida, MDL No. 3006, ruling dated January 15, 2024.
Plaintiffs alleged that Novartis failed to warn that Tasigna, a drug for chronic myeloid leukemia, could cause severe atherosclerotic injuries. They put forward one general causation expert. He opined that Tasigna causes atherosclerosis-related cardiovascular events.
His method began with a documented literature search. He ran a list of search terms through five medical databases and retrieved 1,724 citations. After removing irrelevant and duplicate results, 114 studies remained. Four were randomized clinical trials comparing Tasigna with Gleevec, the baseline leukemia drug. Together they followed 891 Tasigna patients and 614 Gleevec patients.
He pooled the four trials and adjusted each trial's weight for each type of event. Each event type showed an odds ratio of at least 3.0 for Tasigna patients. The lower bound of the 95% confidence interval never fell below 1.86. He also ran a dose-response analysis. When he removed the largest trial, the increased risks remained statistically significant. He then applied each of the nine Bradford Hill factors.
Novartis challenged nearly every step. It argued that the trials lacked a placebo arm and that the largest trial was unblinded. It argued that he ignored the possibility that Gleevec protects against these events. It noted that he did not use a second scientist, as he usually does for publication. It also argued that he failed to rank the Bradford Hill factors.
The Court rejected these points. A placebo arm may be unethical when an effective baseline drug already exists. Attacks on the trials' design went to weight, because they targeted the data rather than his method. The expert had addressed the Gleevec theory directly and cited a study in support. On publication practice, the Court wrote: "But the standards for courtroom testimony are not necessarily the same as for professional publications." Apart from one earlier ruling, the Court knew of no authority requiring a ranking of the factors.
The Court granted the motion in part and denied it in part. The Court barred the expert from offering an opinion that Tasigna causes atherosclerosis characterized as "severe," "rapidly progressive," or "treatment-resistant." In all other respects, the motion was denied, and his general causation opinion was admitted. Novartis's MDL-wide summary judgment motion depended on excluding him, and the Court denied it.
Thompson v. Kinder Morgan Altamont: Reasonable Disagreement Among Epidemiologists
Thompson v. Kinder Morgan Altamont, LLC, U.S. District Court for the District of Utah, Case No. 2:15-cv-00623, ruling dated October 19, 2018.
Clayton Thompson owned property next to a natural gas compressor station that also served as a petroleum storage and freight loading facility. He alleged that pollutants from the station migrated onto his property. In 2012, he was diagnosed with chronic myeloid leukemia, or CML. He sued the companies that had owned or operated the station.
The parties stipulated to resolve general causation first. Plaintiff's expert would testify that benzene, one of the alleged contaminants, is a cause of CML. The Court found him well qualified. He had researched and published on benzene toxicity for more than forty years, for the federal government and in private consulting.
Defendants moved to exclude him. They argued that he relied on studies that did not examine CML or found no significant link, and that he drew conclusions the authors had not drawn. The Court held evidentiary hearings on September 13 and September 27, 2018. Both sides' experts testified at length.
The Court denied the motion. It found that the expert relied on published studies with statistically significant associations between benzene and CML. For studies without such findings, he offered explanations based on accepted statistical principles about rare events. "Scientific consensus is not achieved at once," the Court observed. The literature was not so one-sided that his conclusion lacked support.
The Court framed the dispute as one between two reasonable scientists. Neither expert ignored contrary evidence. They agreed on study design trade-offs and significance levels, and differed on how much methodological choices should temper a study's conclusions. In the Court's words, the dispute was "safely within the bounds of reasonable disagreement between genuine epidemiologists."
Defendants' own expert relied heavily on the review method of the International Agency for Research on Cancer. That body had recently concluded that benzene is associated with CML. The Court found his explanation for disregarding that conclusion unpersuasive.
The Court declined to pick a winner. "Daubert does not direct courts to determine which of two dueling experts is more persuasive. That is the factfinder's province." It found no "unscientific leaps or baseless conjecture" in his review. Defendants remained free to challenge his weighting of the studies at trial.
Hamilton v. Breg: Odds Ratios Too Large to Explain Away
Hamilton v. Breg, Inc., U.S. District Court for the Southern District of Ohio, Eastern Division, Civil Action No. 2:09-CV-146, ruling dated January 24, 2011.
Two young female athletes had knee surgery in 2005 and 2006, at ages 15 and 16. After each surgery, a Breg pain pump infused bupivacaine into the knee joint for at least two days. Both were later diagnosed with chondrolysis. Under the Ohio Products Liability Act, they alleged that the pump caused their injuries.
Plaintiffs retained two epidemiologists to assess the published studies. The first was a full professor of both epidemiology and statistics at UCLA, with a Doctorate of Public Health. He had published more than 300 peer-reviewed papers. The second chaired the Department of Statistical Sciences at Cornell University. He was also a professor of clinical epidemiology and health services research at Cornell's Weill Medical School.
Working independently, the two experts analyzed retrospective cohort studies of shoulder surgery patients. In the first, 12 of 19 shoulders exposed to an intra-articular pain pump developed chondrolysis. None of 158 unexposed shoulders did. Both experts calculated a median unbiased estimated odds ratio of 321. In a second study, 48 of 115 exposed shoulders developed the condition, against none of 281 unexposed. Both calculated an odds ratio of 282. One expert put the figure in context: "This is a larger association than that seen between some of the deadliest exposures and death." They also applied the Bradford Hill criteria.
Breg argued that no study established causation with certainty and that only prospective randomized trials could support the opinion. It faulted the experts for not reviewing patient records, attacked the first study's author as a paid litigation expert, and pointed to the experts' admissions of flaws.
The Court rejected each point. Prospective trials were no longer possible, because surgeons would not risk placing pain pumps in patients' joints. Reviewing clinical records was the study authors' task. "But that is not the role of biostatisticians," the Court wrote. The Court found no evidence that the first study's author was involved in litigation when he did the research. The experts' candor about confounders did not make the studies useless. The strength of the association made the data compelling.
The Court also accepted one expert's refusal to extrapolate his results to every patient in the United States. Limiting conclusions to the data studied was standard practice in his field. On the broader standard, the Court held: "Neither Daubert nor Rule 702 demands absolute evidence of causation before an expert's testimony can be admitted."
Breg's motions to exclude both epidemiologists were denied.
Giles v. Wyeth: A Close Call on a Rare Outcome
Giles v. Wyeth, Inc., U.S. District Court for the Southern District of Illinois, Case No. 04-cv-4245, ruling dated June 18, 2007.
Jeff Giles, a 46-year-old man, died by suicide on October 30, 2002. Several days earlier, his family physician had diagnosed depression and prescribed Effexor, an antidepressant made by Wyeth. He took three pills in the two days before his death. His widow alleged that Effexor was a proximate cause of the suicide and that Wyeth should have warned of the risk.
Plaintiff's expert was a suicidologist with the Psychiatry and Family Medicine Departments at the University of South Carolina School of Medicine. He had investigated and studied thousands of suicides. In the Court's description, "he is a trained epidemiologist." He opined that Effexor caused the suicide. A second expert, a clinician, reached the same conclusion through a differential diagnosis.
The expert gave less weight to randomized trials, which exclude seriously suicidal patients and are not designed to detect rare outcomes. He gave more weight to small studies designed to test whether antidepressants induce akathisia and suicidality. For specific causation, he performed a psychological autopsy covering the decedent's final weeks.
Wyeth relied on a 2006 FDA analysis of 372 randomized trials involving nearly 100,000 people. The FDA found no treatment effect on suicidality in adults overall. Wyeth also argued that studies of suicidal thoughts and attempts say nothing about completed suicide. The Court described Plaintiff's reliance on one age subgroup in the FDA data as "cherry-picked." That subgroup result did not reach statistical significance.
The Court acknowledged that the weight of the epidemiology probably favored Wyeth. Even so, it found peer-reviewed studies on Plaintiff's side with low rates of error. One review of 702 trials and 87,650 patients found an odds ratio of 2.28 for suicide attempts. No large trial had tested the specific hypothesis, so "the Court cannot accept the contention that the FDA's latest study is the last word." It also found no support in the literature for Wyeth's sharp line between suicide attempts and completed suicide.
The Court distinguished Joiner, where none of the studies showed an association. "The differences between this case and Joiner are stark," it wrote. It noted that a number of courts have accepted the psychological autopsy as a method for determining the cause of a suicide.
"Though the point is close," the Court allowed Plaintiff to present general causation to the jury. It called specific causation close as well. Inconsistencies between the two experts were left for cross-examination. Wyeth's motion to exclude was denied in full.
What the Rulings Have in Common
Every featured expert could point to at least one statistically significant result. The Taxotere expert had a reporting ratio with a p-value below .001. The Tasigna meta-analysis had confidence intervals with lower bounds no lower than 1.86. The Thompson and Giles experts relied on significant published studies, and the Hamilton odds ratios ran into the hundreds. The Thompson expert also discussed nonsignificant studies and explained why they did not undercut his opinion.
Where qualifications came up, the courts found them strong. Sanofi did not challenge the Taxotere expert's credentials. The Thompson court found its expert well qualified after four decades of benzene research. The Hamilton court called each of Plaintiffs' experts well qualified.
On method, three themes recur. First, the courts separated attacks on the underlying data from attacks on the expert's method. Unblinded trials in Tasigna, adverse event reports in Taxotere, and possible confounders in Hamilton all went to weight. Second, the experts stated the limits of their evidence. The Taxotere expert's own study acknowledged that reporting data lack a true denominator. The Hamilton experts conceded confounders and possible bias. Neither court held those admissions against them.
Third, the experts engaged with contrary evidence. The Thompson court stressed that neither side's expert ignored adverse studies. In Taxotere, the expert explained why a study reporting a background rate was unreliable.
On fit, the courts accepted reasoned links across closely related evidence. The Taxotere court treated studies of taxanes generally, and of broader eye conditions, as relevant support. The Giles court treated data on suicide attempts as relevant to completed suicide. In each instance, the expert explained the connection.
Practical Points for Attorneys
In each of these rulings, the opinion rested on at least one statistically significant study. Nonsignificant studies served as support in these cases, not as the foundation. Where the expert relied on them, the expert explained why they still fit the opinion.
The experts here stated the limits of their data in their own work. The Taxotere expert's study conceded what a reporting ratio cannot show. The Hamilton experts testified candidly about confounders. In both cases, the courts treated those limits as subjects for cross-examination.
Defendants in these cases often attacked the underlying studies. Tasigna shows the line courts drew. Complaints about unblinded trials and study dropouts went to weight because they targeted the trials, not the expert's use of them.
Dose remains a likely point of attack. The Taxotere court held that general causation did not require a threshold dose under Fifth Circuit law. The expert still had to address dose-response, and he did so with cumulative dose data.
Extrapolation also needed support. Where studies covered a related drug class, outcome, or exposure route, the admitted experts explained the connection rather than assuming it.
Finally, the date of a ruling matters. Thompson, Hamilton, and Giles were decided before the 2023 amendment to Rule 702. The current rule requires the proponent to show that admissibility is more likely than not. Older reasoning remains useful, but it should be read alongside recent authority in the relevant circuit.
The Record So Far
The five rulings span nearly two decades and five kinds of exposure. In each, a court examined how an epidemiologist reached a causation opinion before letting a jury hear it. The Tasigna court drew a line around the scope of the opinion. The Giles court said its decision was close. The Hamilton and Taxotere courts found the method sound and sent the disputes to cross-examination. Each ruling is available by the caption and docket number given above.
Attorneys who want to retain an epidemiology expert can find one on Exlitem. Win Cases With Data, Not Guesswork. The platform holds 52,748+ expert profiles and 245k+ Daubert records across 5,000+ specialties. Search by discipline at exlitem.com.