Patients taking prescription drugs sometimes have bad reactions. Was the reaction caused by the drug, or was it mere coincidence? Was the nausea caused not by the drug, but by the taco for dinner last night?
For subjects enrolled in clinical trials, both the physicians conducting the trials (the “investigators”) and medical monitors overseeing the trials may assess whether an adverse event seems to have been “related” to ingesting the drug. On its face, making that assessment poses no problem; of course companies are interested in knowing whether a new drug appears to be related to an adverse event.
The difficulty arises when plaintiff’s counsel seek to misuse those “relatedness” assessments in product liability litigation. If a subject took a drug and experienced a headache, and the investigator said that the headache was “probably related” to having ingested the drug, plaintiff’s counsel assert that “causality has been established.” The company admits — indeed, the company has proved! — that the drug causes the adverse event, says plaintiff’s counsel.
This is silly, of course. Proving “causality” is tough. If proving causality were easy, why would anyone bother conducting expensive randomized clinical trials? Just give a sick person a drug. If the person gets better, then the drug cures the condition. If the person has a side effect, then the drug causes the side effect. Why bother with science?
This is not quite the way it works. Giving one person a drug and seeing if he or she gets better does not prove that the drug is effective. The FDA requires two positive randomized clinical trials to permit a drug to be labeled as indicated to treat a certain condition.
The converse is also true. If one person takes a drug and experiences an adverse event, that is not proof that the drug caused the adverse event. It might have been the taco, or it might have been coincidence.
That’s why it gets our dander up when plaintiff’s counsel root around through randomized clinical trial data to find examples of an investigator saying that a drug was “probably” or “possibly” related to an adverse event. A “relatedness” assessment is just an investigator’s clinical guess as to whether the adverse event may have something to do with the drug. That assessment is not scientific proof of causation.
Why isn’t it proof of causation?
First, it isn’t science. To establish causation, one must first prove an “association” — people taking a drug experience an adverse event more often than people taking placebo. (And not just “more often,” but more often to a statistically significant level.)
Even once an association exists, “causation” has not yet been proved. Researchers must then “consider whether the association reflects a true cause-effect relationship.” Reference Manual on Scientific Evidence 335-36 (Fed. Jud’l Center 2d ed. 2000). A single “relatedness” assessment is not even proof of an association, let alone the additional evidence needed to suggest a true cause-effect relationship.
Moreover, “relatedness” assessments are often demonstrably wrong. Randomized clinical trials are typically double-blind studies; neither the patient nor the physician knows whether the patient is receiving the test drug or placebo. Physicians often make “relatedness” assessments while the treatment is still blinded — the physician doesn’t know whether the patient actually received the test drug. Any lawyer who practices in this field has seen examples of physicians saying that an adverse event is “definitely” or “probably” related to ingesting the drug, only to learn, after the patient is unblinded, that the patient was not even taking the drug. The patient was taking placebo.
Relatedness assessments do not prove that a drug causes a side effect.
That’s why we were pleased to see the recent decision by Judge Moody in the Accutane MDL. There, plaintiffs insisted (as they always do) that drug manufacturers’ “causality assessments” were an admission that the drug Accutane causes inflammatory bowel disease. The defendants moved to exclude
that evidence under Federal Rules of Evidence 402, 403, and 702. And Judge Moody granted the motion. In re Accutane Product Liability Litigation, MDL 1626, 2007 WL 1288354 (M.D. Fla. May 2, 2007).
Judge Moody wasn’t breaking any new ground here. He held, first, that the “relatedness” assessments were not admissions by a party opponent because they reflect only an assessment of a possible relationship, “not an actual relationship,” between ingesting the drug and experiencing the adverse event. Id. at *3. Judge Moody held, second, that plaintiffs’ expert witnesses could not base their assessment of causality on the companies’ relatedness assessments. “The causality assessments in this case do not contain data which is reliable and upon which an expert opinion of causality can be based.” Id. at *4. And Judge Moody held, third, that the assessments were not admissible to prove that the defendants had received notice that Accutane might be related to inflammatory bowel disease. If “Defendants admit they received notice that some users thought their use of Accutane might be related to their complaints, the reports themselves are unnecessary.” Id. at *5. And, in any event, “to admit the causality assessments for the purpose of establishing notice would be more prejudicial than probative, and would inevitably confuse the jury.” Id.
The only thing odd about the Accutane case is why the defendants were creating the particular relatedness assessments involved in this dispute. These were not the relatedness assessments that are required for subjects enrolled in clinical trials. They were instead relatedness assessments generated for post-marketing reports of adverse events experienced after Accutane was already being marketed. Judge Moody’s opinion says that the defendants made these assessments because the defendants’ foreign affiliates were required to make these assessments for certain European regulatory agencies, and so the defendants made these assessments at all of their companies, foreign and domestic.
Think harder, drug companies. First, assessments that are required only overseas should be made only overseas. The litigation environment may be different in those countries.
Second, assessments that are required overseas can be distorted in American courts because of errors in translation. We have seen counsel assert that French “imputabilite” assessments are “causality” assessments. They are not; they are “imputability” assessments, and they do not purport to assess causality. Errors in translation should not infect American courts.
Finally, why make corporate “relatedness” assessments at all for adverse events reported in the United States? If someone bothers to report an adverse event to a company, can’t the company simply assume that the reporter thought the event was related to the drug? Just assume “relatedness,” and report to the FDA on that basis, rather than generate additional internal documents that can come to haunt you in later litigation.
Relatedness assessments are not causality assessments, and ne’er the twain shall meet.