Wednesday 20 February 2013

Day 117: Trying to shake the Confidence Interval

Today was the first day of the tobacco industry's cross-examination of the plaintiffs' epidemiologist in the trial of the Montreal class action suits.

The companies appear to have learned from their experiences when cross-examining the first few expert witnesses. In November, after a short battle of wits, they quickly threw in the towel when they were unable to best historian Robert Proctor. In January their re-hash of insinuations and insults directed at marketing expert Richard Pollay left the impression of school-yard pugilism more than substantive rebuttal.

But in their response to chemist André Castonguay last week and epidemiologist Jack Siemiatycki today, a new approach is taking shape. Instead of a showy "take-down" or grand reveal of character flaws, the companies now seem aimed at "wearing down" the witness with an onslaught of surprise documents and a series of relentless questions.

The tone has also shifted, and with it the players. The old-style veterans on the tobacco file, Simon Potter and Doug Mitchell, have been moved to the back seats. The Ontario-based counsel for ITL are, for once, sitting quietly. Instead, the limelight has been given to the Quebec members of their teams - Ms. Suzanne Côté (who questioned Mr. Castonguay last week) and Mr. Guy Pratte (who questioned Mr. Siemiatycki today).

If nothing else, their style of substantive steady questions asked in firm-but-polite tones seems better matched to the approach of Justice Riordan.

Bench strength

Although Guy Pratte was the one asking the questions today, it was clear that this was a significant group effort. At his side was Kevin LaRoche, who seemed equally familiar with the questions and where they were heading. More help sat at the rear of the room in the form of hired consultants, additional legal team -- and, it would appear, a visitor from Japan Tobacco's Swiss headquarters.

The pretence of not understanding the science had mostly been thrown off by Mssrs Pratte and LaRoche. Despite taking a pronunciation lesson from the witness over the correct way to say adenocarcinoma (it is not only lawyers who can try to destabilize a verbal opponent!), Mr. Pratte sounded extremely knowledgeable about many of the finer details of this specialty field.  These two men certainly seem to have learned a lot since a few days ago when they said they would not be able to properly represent their client.

The day began with Mr. LaRoche handing over to the plainatiffs and the court officers a CD that contained dozens of documents, none of them known in advance by the witness. This assembled material became the basis for a stream of questions aimed at getting Mr. Siemiatycki to acknowledge that his calculation of the number of Quebecers who had suffered from the four lung diseases that are the subject of the Blais-CQTS case could be an overestimate.

Hacking away at the critical point

Yesterday, Mr. Siemiatycki had explained how he had arrived at the quantity of smoking that was required before a court could presume that it was smoking that caused the disease in question, and not some other factor. Mr. Lespérance had asked a series of questions that allowed him to show how this "critical point" had emerged after a number of logically connected calculations.

Today, Mr. Pratte asked questions that challenged those calculations, and many of the choices Mr. Siemiatycki had made as an inherent part of the epidemiological process.

Why assume the relative risk for smokers for emphysema was the same as for COPD? Why assume a linear dose-response relationship? Why not make separate analyses for each of the lung cancer types? Why use a random effects model instead of a fixed effects model? Why not conduct the same sensitivity analysis for laryngeal cancer and pharyngeal cancer as for lung cancer? Why are other risks for cancer considered multiplicative with risks from smoking, and not additive?

In each case, Mr. Pratte suggested that other choices were available that would have raised the "critical amount" of smoking at which a disease should be linked to tobacco use.

As a seasoned epidemiologist, Mr. Siemiatycki must be used to arguing methodology, and frequently drew on this experience. He carefully refrained from agreeing too readily with statements that were put before him, asking for time to review the whole paper. He was comfortable - and showed it - in talking about the many approaches used in his field, and in the competing views about them. He did not hesitate to agree on the broad lead-up questions, while often (but not always) disappointing Mr. Pratte by not giving him the final answer he was seeking.

It appears Mr. Pratte inhaled

Mr. Pratte repeated an ongoing theme of the defence that population estimates are limited in value and that it  important to know individual smoking histories before coming to any conclusions about any cause of illness.

He artfully used his own smoking history as an example, claiming to have smoked for 10 pack years between the age of 15 and 20. He used his own life to try to show that Mr. Siemiatycki's analysis failed to distinguish between someone like him who was, he maintained. unlikely to become ill from smoking and a more recent quitter.

"Your 10 pack-year history can mean someone who smoked 2 packs a day for 5 years continuously, or someone who packed 1 pack a day for 10 years. It could mean someone who didn't smoke a whole lot, like my mother, who smoked half a pack a day for 20 years. ... If I had an individual person in front of me, I could ask them when they started or when they started and that could make a difference in their risk profile."

"It wouldn't make a huge difference," was the reply.

Herrings (red) and Hetero (geneity).

The one area where Mr. Pratte seemed to make more headway was in his questions to Mr. Siemiatycki about the lack of reference to "heterogeneity" in his analysis, and why reference to this part of the analysis had not been included.

(As I understood it today, if studies are of populations that are too varied, or heterogenous, the results shouldn't be mixed the way that Mr. Siemiatycki had done).

This issue had been included in the criticisms of his report levied by Laurentius Marais. Mr. Siemiatycki appeared ill-prepared to answer these questions today, in part because he had not been directly involved in the mechanics of producing the estimate and in part because "I have not paid attention to Mr. Marais' critiques because I thought they were red herrings."

He said he "would be happy to answer your question, but I would like a bit of time, perhaps a couple of days to think about and reflect and get back to you."  Because of these outstanding questions, Mr. Siemiatycki will return to testify on heterogeneity in March.

Not whether but by how much

If there was a time when tobacco company lawyers cast doubt on the epidemiology that demonstrated that smoking caused lung cancer, it was not echoed in any of the questions put to Mr. Siemiatycki today.

The admissions that Mr. Pratte was trying to wrestle out of Mr Siemiatycki would not have cast doubt on the fact that smoking was responsible for tens of thousands of recent Quebec cancer deaths. They might serve to reduce the estimate, but not by an obviously large amount.


Lots of time today, for example, was spent discussing whether the confidence intervals on estimates of larynx cancer risk should be wider. This would potentially exclude more victims from compensation, but larynx cancers represented less than 2% of the deaths Mr. Siemiatycki counted in his estimate of those "legally attributable" to smoking.

Reflecting on the graphs presented yesterday, it also seems that even if Mr. Pratte succeeded in shifting the slope of the dose-response relationship, or moved the "critical point" over to a higher value, the number of deaths that would be removed from the plaintiffs' case would be very low.

Tomorrow, the cross-examination of Jack Siemiatycki will continue.