We have posted several times in the last few years (like here, here, here, and here) about cases alleging birth defects from maternal SSRI use during pregnancy. Perhaps because of the inherent sympathy for the plaintiff (offspring), it seems that plaintiffs in birth defect cases get extra chances to prove that have enough science to support their claims. Science in this area is undoubtedly complicated. For one thing, while ontogeny does not always recapitulate phylogeny, fetal development of placental mammals (of which humans are a relatively recently evolved species) varies, so extrapolation from animal studies on teratogenic effects is even harder than with studies of direct effects on animals. For another, ethical considerations generally preclude prospective clinical trials designed to study teratogenic effects. FDA’s treatment of drug labeling on the risks of teratogenic effects has reflected some of this dynamic.
Litigations pursuing birth defect claims seem to follow their own dynamic. Despite the oft-cited caution from Rosen v. Ciba-Geigy Corp., 78 F.3d 316, 319 (7th Cir. 1996), that “the courtroom is not the place for scientific guesswork, even of the inspired sort. Law lags science; it does not lead it,” it often seems like cases are pursued before there is science to support them, particularly if specificity of drug, dosage, and injury is going to be required. This may be due to the impact of statutes of limitations. It may also be because some courts are lenient on admissibility of causation evidence and there can be years from filing a case a case until Daubert (or equivalent) motions are decided. Another dynamic seems to be that the only proof as to whether the plaintiff’s mother took the drug during the relevant time can be her word, unless the plaintiff’s expert can argue that the birth defect can itself count as proof of exposure. We also sometimes see that general causation can be a harder concept for courts to understand than it should be.
Looking through our particular lens, we see all of this in K.E. v. Glaxosmithkline LLC, No. 3:14-cv-1294(VAB), 2017 U.S. Dist. LEXIS 13705 (D. Conn. Feb. 1, 2017), a case that came to the correct result—causation opinions excluded—but took an overly charitable route in arriving there. The case alleged that maternal use of the defendant’s SSRI during the first trimester of her pregnancy with plaintiff in 2001 caused him to develop a bicuspid aortic valve and accompanying regurgitation. The plaintiff was diagnosed with a bicuspid aortic valve in September 2010 (after evaluation following his father’s diagnosis with his own heart condition), by which time the drug’s label had been revised to reflect the pregnancy category D—“adverse reaction data from investigational or marketing experience or studies [of the drug] in humans”—and the manufacturer had sent a dear healthcare provider letter describing the basis for the change. Id. at **16-18. The purported prescribing physician testified that he would not have prescribed the drug to a patient whom he suspected was pregnant or recommended it for a pregnant woman based on the label as it was in 2001. Id. at **16-17. We say “purported prescribing physician” because he denied prescribing the drug to plaintiff’s mother until after the plaintiff was born, which is what his records and pharmacy records—but not plaintiff’s mother’s testimony—showed. Id. at **9-12. More on that later. After the bicuspid valve diagnosis, plaintiff’s records suggested that his mother had used the drug during pregnancy and his mother—plaintiff was still a minor—thought that there was a relationship. This was reinforced by a lawyer advertisement on television trolling for plaintiffs and she contracted and signed up with a law firm to pursue a lawsuit over plaintiff’s bicuspid aortic valve by September 2011. Id. at **18-19. Even though the statute of limitations in Connecticut for claims like this was two years, plaintiff waited until late July 2014—close to four years after the diagnosis—to sue. Within a year—pretty fast for such cases—the manufacturer moved to exclude plaintiff’s sole causation expert and for summary judgment on statute of limitations, lack of causation, and other problems with plaintiff’s proof. The decision we are discussing followed more than eighteen months later.
We will save some suspense by noting that the claim was time-barred (as the parties seemed to agree that the mother’s knowledge not the minor plaintiff’s determined accrual) and that almost all the other claims would have failed anyway even if plaintiff had admissible evidence of medical causation. That negligence per se predicated on violating “federal regulations” and “negligent pharmacovigilance,” again based on federal requirements rather than state law duties, were not tossed on their merits need not lead to a rant here. We also do not need to spend time on the analysis that plaintiff could not avoid summary judgment on any claim without a medical causation expert. Instead, we can focus on the Daubert challenge to plaintiff’s expert, a pediatric cardiologist with research experience in congenital heart defects, who opined that maternal use of the drug was a “significant contributing factor” to the plaintiff’s bicuspid valve.
The first challenge was to qualifications and, not surprisingly, it failed. It is hard to exclude professor from a top medical school on a subject that seems fairly relevant to the issues in the case. The court held that his expertise extended to “hypothetical ways that [the drug] might affect embryonic serotonin levels and cause heart defects” even though he had no specific research or writing experience with “serotonin’s role in heart development.” Id. at *29.
The second challenge was to the expert’s general causation opinion. This is where we will take a moment to explain something that seems to be missing: a discussion of the expert’s methodology for forming a general causation opinion. In other words, what did this expert think he needed to establish general causation between the drug and the type of injury and how did he go about analyzing the issue from available evidence? Daubert decisions, including ones cited in K.E., often look to the Bradford Hill criteria as an accepted framework for evaluating the issue of general causation. Like in K.E., the evaluation of epidemiological evidence and the evaluation of biological plausibility tend to take up most of the court’s discussion and both are part of old Sir Austin’s seminal work. The mistake that K.E. made—which did not change the outcome but prolonged the analysis—is that there is no role for biological plausibility unless there is a strong association consistently demonstrated in epidemiological studies. A patchwork causation case may be made through multiple experts—remember, the plaintiff in K.E. had only one—but there is no accepted methodology for evaluating causation, certainly for teratogenic effects, that would let general causation be established with biological plausibility and not epidemiology.
When the plaintiff’s expert failed to offer a reliable opinion that the epidemiology supported general causation, that should have been it. Not just for his general causation opinion, but for the entire case because there is no such thing as an admissible specific causation opinion when general causation has not been established. His opinion on epidemiology failed for the same reason others have failed in SSRI birth defect cases, he did not look at all the relevant studies, specifically seven meta-analyses. Id. at **30-31. While the court did not think this was intentional cherry picking, it rendered his analysis incomplete and unreliable. Id. at **32-33. We cannot tell the strength of association in any of the studies he did consider, if it was statistically significant, or if it was consistent, but it looks like the studies lumped together various cardiac defect rather than bicuspid aortic valves—they are supposed to have three cusps—as they should have to support a general causation opinion, assuming other criteria were also met.
In addition to being moot given the ruling on the epidemiology piece of the expert’s general causation opinion, the court’s consideration of biological plausibility—the area where the expert’s qualifications were also light—was pretty lenient. The expert could not “specify exactly how [the drug] or another SSRI would impact the prenatal environment in a human” or how much serotonin would produce any change relevant to bicuspid valves. Id. at *36. “Potential plaintiffs, however, would be overly burdened by a standard that would disallow any expert testimony on serotonin’s effect on animals because the optimal range of serotonin in pregnant women is unknown.” Id. Daubert imposed a gatekeeping role for a reason, knowing that plaintiffs have the burden of proof. The K.E. court’s response that “the significant limitations of the studies on which [the expert] relies . . . would be best addressed with precautionary instructions to the jury and rigorous cross examination” is misguided when there is a clear analytical gap that should preclude admission. Id. at *37. The same goes for the court’s claim that the jury could “form its own assessment of [the plaintiff’s expert’s] reliability” after hearing from the defense expert. Id. at *38. Assessing reliability is what the court has to do before it admits testimony, as we have said recently.
Based on no admissible opinion on epidemiology and an admissible—but admittedly shaky—opinion on biological plausibility, the court turned to the expert’s specific causation opinion. As we said, the plaintiff’s expert should not have gotten that far. But he did not last long because “the record contains very little evidence that [the plaintiff’s mother] consumed [the drug] while she was pregnant and no evidence from which [the expert] can reasonable infer how much” she took. Id. at *40. There was also no evidence suggesting any use would have been during the time period when it was purportedly biologically plausible that there could have been an effect on aortic valve development. This was enough to preclude any specific causation opinion. The court patiently went through the plaintiff’s arguments that the expert could still offer an admissible specific causation opinion, including the claim that a differential diagnosis does proof of exposure—he could not rule out alternative causes let alone reliably rule in exposure—but there was no rescuing such a speculative opinion. Here, the court noted that there was “too great an analytical gap” between studies suggesting an association—including the epidemiological studies he was already precluded from opining on—and a causation opinion in a case with no real evidence of relevant exposure. Id. at *53. It seems pretty clear that, even with a decade and a half lag from the plaintiff’s birth until the court’s decision, there was too great a gap to allow any reliable opinion of causation. So, while we may think it went on too long—like this post—the result seems correct.