Does glyphosate, the active ingredient in the widely used weedkiller Roundup, cause non-Hodgkin lymphoma? This question is at issue now in a lawsuit in San Francisco Superior Court. Hundreds more claims have been cleared to proceed in a federal multi-district lawsuit.
Much of this litigation is based on a 2015 determination by the International Agency for Research on Cancer, part of the World Health Organization, that glyphosate is a probable human carcinogen. This report has come under heavy criticism, which is not surprising because there's a lot of money at stake.
The IARC classification relied in part on experiments in mice. But is that enough to conclude the weed killer causes cancer in humans? Mice are not people, so probably not.
If it was simple to determine the cause of cancer in humans, scientists would do the right experiment and we'd know the answer pretty quickly.
But it's not simple.
Proving causation in product liability lawsuits
Epidemiology is one of the sciences that provides evidence needed to prove cause and effect in medicine and public health. It is the most important tool for determining whether exposure to a given substance increases the risk of disease. The problem is that it is easy to do it badly, and a bad study is worse than no study at all.
In fact, after a special hearing examining the science on both sides of the glyphosate argument, U.S. District Judge Vince Chhabria called epidemiology "loosey-goosey" and a "highly subjective field." Nonetheless, he concluded that the views on both sides were reasonable and should be heard in court, with the verdict up to a jury.
I have spent much of my working life trying to help figure out why people get cancer. To illustrate how hard it is to prove causality, consider the question: Does smoking cause lung cancer?
Innumerable epidemiological studies since the 1940s have shown a strong association between smoking and lung cancer. But there has never been a randomized trial in humans. In addition, we know from experimental studies that smoking rats don't get lung cancer.
For years, Big Tobacco dismissed observational studies in people (epidemiology) with the mantra that "association is not causation," and avoided regulation. The scientific community was intimidated by this strategy for far too long.
Eventually, the studies accumulated to the point that the association was overwhelming, and cause and effect could not be denied.
There are two main types of epidemiological study designs: cohort and case-control. In a cohort study, a large group of people – some smokers, some not – are followed over the years to see who gets sick.
In a case-control study, a group of lung cancer patients (perhaps several hundred) are asked about their smoking history, along with an equal number of people without lung cancer.
Invariably, in cohort study after cohort study, smokers got sicker from heart disease, lung cancer and many other maladies over time. In most of these studies, scientists did their best to take account of other differences between smokers and non-smokers, so as to isolate the effect of smoking.
Also invariably, in case-control studies patients with lung cancer were much more likely to have been smokers than people in the general population.
When scientists are asked for a definition of proof, most of them use criteria such as "reproducibility" and "statistical significance" and "plausibility." But who decides whether each of these criteria has been met?
The answer is a panel of experts. It is unsettling to most scientists to hear that "proof" can only be defined as "a consensus of experts," but this is true from physics to bird-watching. And what has been proven can later be unproven with new experts and/or new evidence.
Who chooses the experts? They include panels convened by the National Academies of Sciences, or advisory boards of professional societies such as the American College of Cardiology. The makeup of these panels can be challenged, and of course, people can choose to ignore the "experts" and believe what they want.