In the Paxil studies, many of the "designed characteristics have been such that they would minimize or make it more difficult to detect an increased risk," she said. "And despite that, these studies have shown consistency in showing an increased risk of cardiac malformations associated with first trimester Paroxetine exposure."
"The pressure is always against the ability to detect increased risk in the way these studies are designed," Kramer said. "And, yet, despite that, we are seeing consistently elevated risks associated with Paxil, which is very, very important, very compelling, and very alarming actually."
Kramer described the difference between association and causation as meaning that a single study with a finding of an elevated risk of birth defects would only show an association. "When you have a body of literature which shows through multiple studies consistently elevated findings, then you move from association in one study to causation, that this factor causes the disease," she told the jury.
During closing arguments on October 8, 2009, Tracey told the jury that, "Defense lawyers can't stand the word 'causal.'"
"Causal" is the "kiss of death" for a defense lawyer, he said, because they know that is one of the questions the jury will be asked.
"The second question you are going to be asked," he told the jury, is "Do you find that Michelle David's ingestion of defendant's drug Paxil was a factual cause in bringing about the heart defects?"
Epidemiology 101
While testifying, Kramer explained what is meant by relative risks and confidence intervals. "Our real interest in epidemiology is to measure rates of disease and excess risk," she said. "But we also want to know really how precise is this measure."
"And the precision of this measure is very much tied to the size of the population that you are studying and the number of exposed people," she explained.
"In other words," she said, "if we were to go into a large population and do the same study a hundred times, how many times out of a hundred would we find the same exact answer?"
"It is similar to tossing a coin," she noted. "If you are looking at the proportion of heads and tails in a coin toss, and you toss that coin a thousand times ... you are going to come up with that 50/50 proportion pretty much all the time."
"That's a very precise answer," she pointed out.
"So if you think about it that way," Kramer said, "the larger the sample size, the larger the number of people that you study, the more precise your study estimate of that relative risk is."
"And we estimate the precision of this relative risk by calculating something called confidence interval," she told the jury. "If you were to repeat this study, let's say 95 times out of a hundred, what would that range be?"
For instance, where the relative risk in a study is 2, and they calculate statistically a 95 percent confidence interval with a range of between 1.5 and 2.5, the actual relative risk would fall somewhere in this range. That "means 95 trials out of a hundred would generate results in this range," Kramer stated.
A test that is not statistically significant should not be discarded, she said. The "practice of statistical significance testing has been very much rejected in epidemiology because it was never developed really to study health or biomedical or human health problems."
