Monday, 10 March 2014

Day 213: Not your "average" kind of witness

Good news should travel fast!

Justice Robert Mongeon issued his decision last Friday to the tobacco industry's constitutional challenge to the the law which underpins the Quebec government lawsuit. Five long months have passed since he listened to a week of arguments

Spoiler alert! Justice Mongeon's ruling is not yet available on Quebec's SOQUIJ database of rulings. But why wait to read the details before knowing that those who were celebrating on the weekend were the Quebec government's legal team, on whose side Justice Mongeon came down firmly.

He ruled that the provisions of the Tobacco-related Damages and Health Care Costs Recovery Act, taken in isolation or as a whole, did not offend either Article 23 (right to a fair trial) or Article 6 (property rights) of Quebec's Charter of Human Rights and Freedoms. The law was "clear and unambiguous" he wrote, in the way it eliminated a limitation period and was therefore consistent with the Charter.

Oh, my goodness. It's already March, the clocks have been pushed forward and we are entering what may be the last witness-sprint in the Montreal Tobacco Trials. Is the end really coming into sight?

Certainly, there was an atmosphere of end-of-term intensity in the courtroom when the lawyers for the 5 parties re-assembled this morning. (The regular one-week-a-month break came after a very reduced sitting schedule for February).

All sides were there in full force for the beginning of one of the more strategically important moments at this trial: the assault on the plaintiff's expert epidemiologist.

Last year, Mr. Jack Siemiatycki presented his estimate of the number of Quebec smokers  (110,282 people) who have fallen victim to one of the the four diseases included in the Blais class action. He calculated the number of years one would have to smoke at the rate of a pack-a-day before such diseases could be considered  - more likely than not - to have been caused by smoking (3 to 5 pack-years).

Justice Riordan seemed quite accepting of these numbers, and Dr. Siemiatycki's opinion clearly influenced his ruling for a new definition of class members in the Blais case.

Something to aim at

Mr. Siemiatycki's report feels like the foundation of the Blais disease-case. It quantifies the harm, sets a threshold for the number of deaths that can be "legally attributable" to smoking, and establishes the rationale for treating claimants as a population, instead of just an aggregation of individuals.

That's why for the tobacco industry's side, it is THE bulls-eye to aim for. A knock-down punch on Mr. Siemiatycki's analysis could destroy the logic of collective recovery in this trial, or at least re-route it towards Engle-style individual claims.

Three skilled witnesses are lined up to take a goat the report over the next two weeks of trial. The first, today's witness, is Laurentius Marais. Watching him closely from the public gallery on the defence side of the court were the other two industry experts (Mssrs. Kenneth Mundt and Bertram Price). Mr. Jack Siemiatycki sat ahead of the bar with the plaintiffs' team.

Marthinus Laurentius Neethling Marais

Laurentius Marais in 2000 testifying for the Bush Campaign
The voir-dire to qualify Mr. Marais did not take long. Mr. Guy Pratte (counsel to JTI-Macdonald, who is Mr. Marais' client in this case), gave a very slender introduction - in spite of or perhaps because of the technical difficulties that unusually hampered both sound and visual displays this morning.

Mr. Marais described his move from his home town of Stellenbosch in South Africa to the United States over 40 years ago, leaving for graduate studies in California. (Stellenbosch is coincidentally also the home of the Ruperts and their Rembrandt/ Rothmans empire).

Four decades of study (Stanford, mathematics and business), teaching (University of Chicago, Associate Professor) and consulting in the United States have erased all but small traces of the distinctive South African accent.

From his base in Wyoming, Mr. Marais has spent the past 20 years engaged by various industries in their litigation or regulatory work. He is a senior executive at William E. Wecker, which has a broad if frequently controversial clientele.

Mr. Pratte invited his witness to explain how it was that a statistician who had not published on lung diseases was in a good position to critique an epidemiologist.. "Applied statistical methods are at the core of what Dr. Siemiatycki attempts to do." And as for his experience in epidemiology, he curiously cited his work opposing legal claims by "haemophiliacs in the US who had become infected through medicines made from human plasma."

Although Mr. Pratte did not ask him about it, the most reported of Mr. Marais' consultancy experience is likely his work to support the Republican Party in the court review of the 2000 Florida vote. During that trial Mr. Marais was described by a lawyer for the Al Gore/Democratic Party as "a man who makes a living out of killing the testimony of other statisticians at the behest of evil interests."

Mr. Johnston, during his brief cross-examination on the voir-dire, also made a rare reference to the concept of evil. As he has done with other witnesses, he asked Mr. Marais to consider his testimony in light of the ethical standards set by his profession. (In this case the American Statistical Association, Exhibit 1697).
Statistical tools and methods, as with many other technologies, can be employed either for social good or evil. The professionalism encouraged by these guidelines is predicated on their use in socially responsible pursuits by morally responsible societies, governments, and employers. Where the end purpose of a statistical application is itself morally reprehensible, statistical professionalism ceases to have ethical worth.
Mr. Marais sounded a little confused about the existence of such a guideline, but did not shrink from the language. "As a general statement, I can't disagree with it," Mr. Marais said, adding that he felt that his work was consistent with that approach.

There was more confusion over how many times Mr. Marais had testified on behalf of tobacco companies. "Half a dozen to a dozen" times was his answer today. But when Mr. Johnston pointed out that ten years ago he said it was "10 to 15" times, he took refuge in statistics: "there's an overlap between those ranges!"

There was no dissembling about his views on causality, however. "I accept that it is a cause of many cancers, that it is a cause of cardiovascular disease and other respiratory diseases."  

At the request of Mr. Pratte, and with only a limp objection from Mr. Johnston, Justice Riordan qualified Mr. Marais as "an expert in applied statistics, including in the use of biostatistical and epidemiological data and methods to draw conclusions as to the nature and extent of the relationship between and exposure and its health effects." 

Mr. Marais' Expert Opinion

The dozens of expert opinions filed in this case are often a hundred or more pages long, and I confess that with the passage of time my memory is more than a little blurred. Nonetheless, I would hazard that Mr. Marais' (Exhibit 40549) is, with the possible exception of Mr. Duch, the most vitriolic and sharply worded of any of the defence critiques.
"This fundamental logical defect of the method as well as Dr. Siemiatycki' s flawed implementation of it-reasons ..."
"What Dr. Siemiatycki proffers is not a validated method .... It is, rather, an untested amalgam of statistical procedures that is incapable of producing statistically valid estimates"
"The estimates produced by Dr. Siemiatycki's method are fatally infected with the defects of the method and cannot be cured of them."
I noted that his oral testimony today was much less acidic. Perhaps it was the tone set by Mr. Pratte, whose trademark style seems rooted in the gentillesse of bygone times.  This courtly refinement seems a good match with Justice Riordan's expectations, and thus a good ground for Mr. Pratte to set as he prepared his witness to take aim at the Siemiatycky bulls-eye. 

One volley after another, Mr. Marais took aim.

Just as Mr. Siemiatycki had presented his conclusions in a step-by-step progression, Mr. Marais used a building-block approach to trying to dismantle his conclusion. His challenges were delivered in well-constructed full sentences, delivered at a judge-friendly slow pace.

Justice Riordan's contributions added to the tone of intense but respectful attention. His occasional questions over the day showed that he listening carefully but critically -- occasionally challenging Mr. Marais' characterization, or asking for clarification on subtle points.

There were a few clear patterns that seemed to emerge from these varying complaints. The flaws were fatal: for differing reasons, Mr. Siemiatycki's calculations could not be re-done to correct flaws or accommodate criticism. And the population-approach taken by the epidemiologist could not be reconciled with the statistician's understanding of the individual experience of smoking or getting sick.

Many arrows in his quiver

Mr. Marais may have agreed that "cigarettes caused disease," but he responded to Mr. Pratte's questions by distinguishing between "general causation" in a population, and "specific causation" in an individual. He frequently returned to the theme of assessing individual/specific harm as a barrier to accepting Mr. Siemiatycki's approach.

He said, for example, that there was no scientific basis to assess disease causation in an individual by appealing to a calculated "critical amount" of smoking beyond which the "probability was greater than 50%." 

In any event, Mr. Siemiatcyki's calculation of such a "critical amount"  had not been properly done. His meta-analysis had overlooked too much data. As a result, one could not say that any particular amount of smoking created a threshold limit to infer causation.

Mr. Siemiatycki had erred by lumping smokers together. There was too much "heterogeneity" in the population for averages to be appropriate. (The idea of an average seemed an unlikely anathema to this statistician!)

Mr. Marais introduced the concept of the wrongness of measuring the height of Quebecers, calculating the average height and then "assigning that average back to each individual member," a comparison he returned to over the day. 

Confidence intervals, or the lack of them, was another repeated criticism. The court was shown his "illustrative example" of how the confidence interval must not overlap a threshold value before one could safely conclude, on a balance of probabilities, that a statistical result could be inferred. 

If this had been properly taken into consideration by Mr. Siemiatycki, the critical amount for larynx   cancer would grow from 4 pack years to 25.

In addition to these "sampling error" concerns, Mr. Marais also found systematic (or non-stochastic) errors.

He said it was wrong to treat the four major types of lung cancer as a single disease. Individuals came with only one of those conditions, he pointed out.

He felt that the determination of the "critical amount" of smoking was one that was not "accurate" in statistical terms, and that therefore "anything that flows from it, like the calculation of the number of class members, inherits that problem." 

He acknowledged that pack-years was a frequently used metric, but did not think it properly measured intensity nor duration of smoking. He pointed to the different risk profiles of smokers who quit decades ago versus smokers who quit more recently, or who had not quit. "For smokers who quit by age 30 – even by age 35 – after 10 or 20 or 30 years, their level of risk for class diseases approaches that of people who never smoked." 

Nor can these groups be averaged. "There an be no person who personifies the average smoker. It is not possible for a person to be both a continuing current smoker and one who quit many years ago."

He felt Dr.Siemiatycki had erred when using a "multiplicative method" to consider combined risks (i.e. smoking and drinking, or smoking and being exposed to radon), instead of an "additive method." This second approach would increase the attribution to other causes (again HPV was cited!), while with multiplicative approaches "smoking always doubles the risk."  

By late afternoon, the criticisms were entering territories usually only occupied by the methodologically obsessed. Dr. Siemiatckyi had apparently erred in assigning "midpoints" when using a study whose results had been presented in ranges. And we are not done .... tomorrow Mr. Pratte will invite further criticisms from his witness.

Mr. Marais is scheduled to testify again tomorrow and Wednesday.