Tuesday, 19 March 2013

Day 128: Methods and methodology

In the competitive theatre of the Montreal tobacco trial, there are occasionally dramatic shifts between scenes. Today was one of those days. Yesterday's squabbling about old grievances gave way to today's high-toned and civil exchange about epidemiological methods.

The curtain on today's episode did not came up until the afternoon, as the morning's hearing had been sacrificed to an unspecified scheduling conflict. When it did open, the courtroom setting had been tweaked for the re-appearance of Montreal's noted epidemiologist, Jack Siemiatycki.

The counsel for Canada's smallest multinational tobacco company, JTI-Macdonald, had moved to the defence's front seat, which is on Justice Riodran's left hand side and beside the witness. Sitting beside Guy Pratte at the front desk was Kevin LaRoche, who seems responsible for preparing for scientific witnesses for JTI.

With Mssrs Pratte and LaRoche in charge of the day, I expected a smooth ride. These players don't go in for the theatrics, pissing matches and games of silly bugger that some of their colleagues excel at. But even against my relatively high expectations for the day, I was pleasantly surprised.

It was these two men who would lead the remaining cross-examination of Mr. Siemiatycki, a witness whose report is very damaging for the companies. The report (Exhibit 1426.1) concludes that in recent years the number of Quebecers made sick as a result of smoking these companies' brands is four times the population of Outremont:

"From 1995 to 2006, there were 110,282 Quebec residents who suffered from one of the four diseases and who would have had probability of causation greater than 50% that the cause was cigarette smoking."

Epidemiology 401

In trying to challenge the methodology used by Mr. Siemiatycki, this duo had gone to obvious efforts to bone-up on the content, to refine the issues and to present their challenges in ways understandable to judges (and bloggers!) who have more limited expertise in quantitative methods.

They did a good job!

This difficult challenge may not have left room for any distractions that come with the bag of tricks that has been used when questioning other plaintiff's expert witnesses. Jack Siemiatycki was spared the mockery and challenges to his judgement and scholarship that were experienced by Robert Proctor, Richard Pollay and André Castonguay.

Instead, most of the day was spent with Guy Pratte leading the witness through a series of finely sequenced questions aimed at getting the epidemiologist to acknowledge that the methodological choices he had made would have affected the reliability of his results.

During the first couple of hours of this effort, I felt like I had accidentally dropped in on an upper-year lecture on statistical methods. Mr. Pratte was at the head of the class, explaining concepts underlying the measurement of differences between studies or the attribution of those differences.  Q value! I squared! 

The pedagogical nature of the exercise was emphasized by the introduction of extracts from text-books, and by the illustrations offered when these concepts were applied to Mr. Siemiatycki's calculations. (Exhibits 40042.1, .... 40042.5)

Mr. Siemiatycki responded very politely throughout this process. He either agreed with Mr. Pratte, or did not let his disagreement get in the way of the groundwork that the lawyer was making. He offered the occasional conceptual reservation, but was quick to express his willingness "for the sake of the argument to along with the idea." 

In less time than a standard university lecture period,  Mr. Pratte had finished the ground work and began to present his central challenge to Mr. Siemiatycki's conclusions: "Would you agree that the extent of heterogeneity would impact the reliability of your summary data [about lung cancer]?"  

Mr. Siemiatycki quite agreeably replied. "As a generic statement, yes," he said, then added "But in particular circumstances it might be moot." 

It was not until some time later (and after a brief flurry of excitement when Mr. Siemiatycki disclosed a calculation error) that the witness began to politely pour cold water on Mr. Pratte's argument.

"There are many reasons why different studies may vary from one to another," he explained. "The most important reasons are probably methodological." 

The issues raised did not affect his overall conclusions, he said. "I became convinced that no matter what level of heterogeneity there might be among the studies, the range of values from all of the studies was so far off the chart for what we usually see in terms of the magnitude of relative risk and dose response relationships, that it would have little impact on the final results."

He outlined in numerical terms the impact that would be felt and then provided a more memorable way to understand the issues. "To answer a question about whether the Himalayas are higher than the Alps, you don’t need the most specific measure possible."

He said that the difference would be at the "margins" of the smoking population, with people who have only a handful of pack-years smoking experience. "All of these arguments ... are trivial because nobody smokes that little."

Mr. Pratte tried to introduce other methodological concerns -- like the failure to include "prediction intervals," or possible differences in the types of cigarettes smoked or the population studied. But these issues were on Mr. Siemiatycki's home turf, and easily rebuffed. With confidence and eloquence, Mr. Siemiatycki defended the appropriateness of drawing inferences in the way he had done.

In the face of such well articulated knowledge, Mr. Pratte was at a disadvantage. He temporarily relied on some lawyerly questions -- "you can't exclude...", "you didn't investigate..." but soon announced that he had no more questions to ask.

There were two other quick rounds of questions from Simon Potter (PMI/Rothmans, Benson and Hedges) and Allan Coleman (Imperial Tobacco), both focused on assumptions of linearity in the models.

At the end of the day, as the plaintiffs' were able to respond to the cross-examination, there were already clear signals from Justice Riordan that he did not need to or want to hear any more. Nonetheless, André Lespérance wanted his witness to clarify some issues.

He focused on the core of the methodological concerns. What if Mr. Siemiatycki approach miscalculated how much someone had to smoke before tobacco use was the likely cause of disease?  What if there were a different "critical amount."?

Mr. Lespérance put it this way: "How critical is the critical amount to the bottom line?"

"Not very critical," replied Mr. Siemiatycki. "Estimates can be made with different models and different assumptions, but once we are in the range of pack years from 2 to 12, we will end up with the same number."  He explained that the vast number of lung cancers were found among people who smoked for 20 or more pack-years. "The average pack years for lung cancer cases is 50."

Nor were other causes of cancer - including asbestos and alcohol - likely to greatly affect his conclusions. "It is like Mount Everest compared to Mount Royal - which one can obscure the other one." 

In the months to come, the epidemiologists and statisticians hired by the companies will present their reports, and Mr. Siemiatycki is expected to again offer his insights at that time. If today is something to go by, it is a day to look forward to.

Tomorrow and Thursday the trial will hear from the plaintiffs last expert witness, Dr. Juan Negrete. He will testify about addiction.