Kelly Havner was born with a limb-reduction defect: the fingers of her right hand did not form. Her mother took Bendectin, a prescription drug manufactured and marketed by Merrell Dow, during the period of pregnancy when Kelly’s limbs were developing. The Havners alleged that Bendectin caused the defect and sued on negligence, defective-design, and defective-marketing theories.
At trial, the Havners offered epidemiological studies, animal studies, in-vitro studies, chemical-structure evidence, and testimony from five experts. The jury awarded $3.75 million in actual damages and $30 million in punitive damages, later reduced by the trial court to $15 million. A court-of-appeals panel initially rendered a take-nothing judgment for Merrell Dow for lack of causation evidence. Sitting en banc, however, the court of appeals affirmed the actual-damages award while reversing the punitive damages. The Texas Supreme Court granted review and addressed only Merrell Dow’s no-evidence challenge to causation.
Issue #1
Whether a court conducting no-evidence review may examine the scientific reliability of admitted expert testimony rather than accept the expert’s conclusory assertion of causation as evidence.
Holding
Yes. Scientifically unreliable expert testimony is legally no evidence, even if it was admitted at trial and even if the expert states an opinion in terms of reasonable medical probability.
Reasoning
Texas no-evidence review requires a court to consider the evidence favorably to the verdict, but it does not require the court to treat labels or conclusory language as probative proof. An expert’s use of phrases such as “reasonable medical certainty” cannot substitute for a reliable scientific basis. Otherwise, expert testimony would effectively be insulated from meaningful sufficiency review.
The Court held that the reliability principles governing admission of scientific evidence under Texas Rule of Evidence 702 and Robinson also help determine whether an admitted expert opinion has probative force. Courts may examine whether the underlying data are reliable and whether the expert used a sound method to draw conclusions from those data.
The relevant Robinson factors include testability, the degree of subjective interpretation, peer review and publication, the potential rate of error, general acceptance in the relevant scientific community, and nonjudicial uses of the method. An opinion founded on unreliable data, or on an unjustified inference from otherwise sound data, is not evidence that can support a verdict.
Issue #2
Whether epidemiological evidence can support causation in a toxic-tort case, and what a plaintiff must show when relying on that evidence.
Holding
Properly designed and executed epidemiological studies may support causation, but a claimant ordinarily must show a statistically significant relative risk greater than 2.0, together with reliable evidence that the claimant is comparable to the studied population and that other plausible causes have been reasonably addressed.
Reasoning
The Court distinguished general causation—whether a substance can cause a condition in the population—from specific causation—whether it caused the particular plaintiff’s injury. Epidemiological studies identify associations in populations; they ordinarily cannot identify the actual cause of one person’s injury. The legal system may nevertheless permit an inference of individual causation from reliable population evidence when the inference satisfies the preponderance standard.
A relative risk greater than 2.0 reflects more than a doubling of risk. As an illustration, if an injury naturally occurs in six of every 1,000 unexposed people but in more than twelve of every 1,000 exposed people, more than half of the cases among the exposed population may statistically be attributable to the exposure. The Court concluded that this approach has a rational relationship to the requirement that causation be more likely than not.
The Court rejected any mechanical rule that a single study reporting a relative risk above 2.0 automatically proves causation. Epidemiological results must be statistically significant, generally at the conventionally accepted 95% confidence level, and courts must consider study design, sample size, bias, confounding variables, biological plausibility, consistency, dose-response relationships, peer review, replication, and other accepted causal criteria.
A plaintiff must also connect population evidence to the individual case. That ordinarily requires proof of exposure to the same substance, a comparable or greater dose, exposure before injury, a timing of injury consistent with the studies, and reasonable exclusion of other plausible causes when they can be negated. The Court left open whether epidemiological evidence showing less than a doubling of risk, combined with other credible and scientifically reliable proof, could ever suffice.
Issue #3
Whether the Havners presented legally sufficient scientifically reliable evidence that Bendectin caused Kelly Havner’s limb-reduction defect.
Holding
No. The epidemiological, animal, in-vitro, and specific-causation testimony did not provide a scientifically reliable basis for finding that Bendectin caused limb-reduction defects generally or Kelly Havner’s defect specifically.
Reasoning
The epidemiological studies cited by the Havners did not demonstrate both a statistically significant association and a more-than-doubled risk of limb-reduction defects. Many reported relative risks below 2.0, confidence intervals that included 1.0, or both. A confidence interval including 1.0 means the study is statistically inconclusive at the stated confidence level because it is consistent with no increased risk.
The Havners’ experts attempted to reanalyze selected data to generate higher risk estimates, but the record did not establish essential features of those analyses, including adequate confidence levels or confidence intervals. The experts’ litigation-driven reanalyses had not been published, peer reviewed, or replicated, even though more than thirty published peer-reviewed studies had found no association between Bendectin and birth defects. The Court concluded that the experts could not selectively reinterpret studies in a manner that sound scientific methodology would not accept.
The animal studies did not establish human causation. Some involved extraordinarily high doses that were not meaningfully connected to the dosage Marilyn Havner took, and the experts did not explain a scientifically valid extrapolation from those animal doses to human pregnancy. Animal data may justify further research, but they could not, without confirming human evidence, prove that Bendectin more likely than not caused human limb defects.
The in-vitro studies showed that doxylamine succinate could affect animal cells or tissue in laboratory conditions. But evidence that a substance can affect isolated cells is only a starting point for scientific inquiry; it does not establish that the substance causes birth defects in whole human beings. The expert offered no reliable bridge from the laboratory findings to his conclusion that Bendectin was a human teratogen.
Dr. Palmer’s opinion that Bendectin specifically caused Kelly Havner’s defect rested on the same unreliable epidemiological and animal evidence. His evidence about timing of exposure and the drug’s ability to cross the placenta did not cure the absence of reliable proof that Bendectin causes limb-reduction defects in humans. His conclusion therefore overstated its scientific predicates and amounted to personal opinion rather than legally sufficient causation evidence.