When You Have to Prove a Doubling of the Risk Ask "For Whom and During What Time Period?"
Despite the fact that the Texas Supreme Court in Merrell Dow Pharmaceuticals, Inc. v. Havner wrote "[w]e do not hold, however, that a relative risk of more than 2.0 is a litmus test..." many lawyers and some courts believe that Havner (and even Daubert) require that plaintiff establish epidemiological evidence of a "doubling of the risk" before she can establish so-called specific causation. In my view all those two courts ever said was that if you wanted to play the game of epidemiological causal inference then you had to play by epidemiology's rules; and, furthermore, that if all you had was probabilistic evidence then that evidence had better show that defendant's product probably did it.
Whatever the interpretation, we mass tort lawyers often wind up fighting over whether there has been a doubling of the risk. One thing we're just beginning to fight over is "when do you measure the risk?" Take for instance the Women's Health Initiative and hormone replacement therapy (HRT).
Depending on when you decide to look HRT either caused an 80% increased risk (after one year of treatment) of coronary disease or a 30% decreased risk (after five years) of coronary disease. So how do you choose which risk is the "real" one?
In what promises to become an epidemiology blog posted for free at Epidemiology is a copy of "The Hazards of Hazard Ratios". In it the author makes the point that hazard ratios, an approximation of risk ratios, often vary over time and may be subject to biases, as when, hypothetically, over time the exquisitely sensitive, by virtue of suffering the malady earlier, manage to be deselected from subsequent years' calculations.
The phenomenon is well known to those doing benzene litigation. Had the study of the Pliofilm workers been done today the very same cohort would demonstrate a relative risk less than 2.0 quickly trending towards 1.0. The obvious retort is that "latency" somehow is responsible but that doesn't explain the fact that no member of the cohort has developed leukemia in decades. The most sensible answer is that those susceptible of getting it get it and those who aren't don't just fail to get it - they can't.
In other words, the distinction in toxic torts between "general causation" and "specific causation" is likely often (if not usually) a false one. Thanks to the laws of physics, bullets can reliably be said to be a general cause of bullet holes in people since all are susceptible to the effect. The same cannot, however, be said of benzene or HRT. Apparently they only increase the risk (perhaps to 1.0) in those people who, due to genetics, epigenetics, microbiota, environment or some other factor(s), are primed to produce the effect while simultaneously imposing an increase of 0.0 on everyone else.
So how, other than flawed post hoc "reasoning", do we determine whose injury was caused but for an exposure? Biomarkers have clearly failed of their promise. Now what? The way things have been going I bet it'll look something like this: "Discovering Graphical Granger Causality Using the Truncating Lasso Penalty" but maybe we'll get lucky and it'll look more like: "Causal Diagrams and Change Variable". In the meantime plaintiffs without biomarkers need to do a better job of demonstrating how they are like those in the time interval with the biggest risk.
