Chris Snowdon, author of Velvet Glove, Iron Fist, kicks off the Colloquium About The Cigarette Hypothesis:.
Smoking and lung cancer: The case for the prosecution
I’m as interested as anyone to see what crops up in this discussion. ‘Smoking causes lung cancer’ is regularly cited as one of the strongest findings in epidemiology, and so it is. It is certainly one of the most significant (in both the statistical and conventional sense).
Like most scientific findings, "the smoking theory" (as the tobacco industry used to call it) is taken on trust by those of us outside the field. I was not even born when most of the major studies were published. When writing Velvet Glove, Iron Fist, I sought out every study on secondhand smoke and lung cancer. I can’t say I did the same for primary smoking and lung cancer but I read enough for me to be satisfied that smoking raises the risk of lung cancer for a lifetime smoker by a magnitude of 10 to 20 (somewhat less for some ethnic groups, particularly East Asians).
For most of us, it is proven beyond any reasonable doubt, but that does not put it beyond debate. As long as there are people who question a claim, the debate is not over. Debate is healthy and necessary. Gallileo did not persuade people of his theory by calling them flat-earthers. Darwin did not persuaded people of his theories by calling them deniers. And so I am not going enter this debate by appealing to authority ("every expert agrees that…") or with weasel words ("everyone knows that…"). Nor do I need to. I believe the evidence speaks for itself.
A suitable place to kick things off is Austin Bradford Hill’s famous speech of 1965, which set out what was effectively a check-list of criteria to be met by epidemiologists before they cried "There! That’s the man!". Hill was the co-author of one of the earliest smoking-lung cancer studies in 1950, to which we will doubtless return, and he was acutely aware that statistical correlations were common, but genuine causative associations were not. I’d like to think that contemporary epidemiologists wake on Christmas Eve every year to see the ghost of Hill at the end of their bed, but they probably sleep like babies (the swine).
Anyway, here are his criteria and how they relate to smoking-lung cancer:
1. Strength: Hill gave the example of chimney sweeps and scrotal cancer which found a relative risk (RR) of 200 (ie. they were 200 times more likely to get this rare disease). Such a relative risk was so large that it required no epidemiological study. It was observable to the naked eye, as it were—as obvious as finding the association between mining and coughing, or being a woman and enjoying Sex in the City. The lung cancer-smoking link (from hereafter "the smoking theory" just to wind up any ASH supporters) is not quite as large as that—hardly any are—but it is generally in the region of 5 to 20, always greater than 2 to 3 and sometimes as high as 50. By any standard, this is a strong association.
2. Consistency: As Hill said: "Has it been repeatedly observed by different persons, in different places, circumstances and times?" Clearly the answer is yes. The smoking theory has been demonstrated again and again all over the world since the 1930s.
3. Specificity: In other words, does the disease only affect the exposed group? The fact that there is no malaria without mosquitoes supports the theory that mosquitoes spread malaria. Asbestosis is entirely specific to people who have been exposed to asbestos. Is there the same specificity with smoking and lung cancer? The answer is no. People developed lung cancer before smoking arrived in Europe and nonsmokers still get lung cancer in reasonably large numbers. Other factors are involved—other forms of smoke, radon and several other risk factors. But, as Hill said, people can get scrotal cancer without being chimney sweeps. He continued: "If other causes of death are raised 10, 20 or even 50% in smokers whereas cancer of the lung is raised 900 – 1000% we have specificity – a specificity in the magnitude of the association."
4. Temporality: In other words, cause and effect. Does smoking cause lung cancer or do smokers happen to be people who put themselves at risk of lung cancer more than nonsmokers. This is a valid question that less scrupulous epidemiologists fail to ask. We know today that smokers are more likely to be in lower socio-economic groups. This is a major confounding factor. The fact that lower socio-economic groups are also more likely to end up in prison and are more likely to have a baby die in the first 12 months of life does not prevent junk scientists claiming that secondhand smoke "causes" criminality or cot death. But these associations are very weak whereas the smoking theory is strong. If there was a more significant risk factor for lung cancer that it associated with smoking, but is not smoking, we need to hear what it is. The suggestions put forward—such as vehicle exhaust, asbestos exposure or pollution—are not specific enough to smokers to explain the statistical association between smoking and lung cancer.
5. Biological gradient: Does the risk rise with the number of cigarettes smoked or the number of years of smoking. With a very good degree of consistency, studies have shown the answer to be yes.
6. Plausibility: Is it plausible that filling the lungs with smoke over a hundred times a day could damage them? I would say so. If not that, then what? Sometimes I hear smokers’ rights people disputing the smoking theory while blaming car exhaust fumes. This seems to me to be inconsistent. Either breathing fumes, toxins and carcinogens into the lungs is dangerous or it is not. There is almost certainly a threshold below which the human body can filter and tolerate these fumes (which makes passive smoking and multiple chemical sensitivity less plausible as serious risks), but directly breathing smoke in must be well above that.
7. Coherence: Does it fit the overall pattern? Again, yes. The lung cancer rate rose first amongst men because they were the first to smoke and it rose later amongst women when they started to smoke. It is highest in regions and countries where people smoke and lowest where they do not. In Sweden, the use of snus has led to Europe’s lowest smoking rate and also the lowest lung cancer rate. Is there a more plausible explanation for this than that smoking causes lung cancer?
8. Experiment: This is the most difficult aspect of the smoking theory since lung cancer takes decades to develop and is therefore difficult to demonstrate in a laboratory experiment. Scientists have successfully made animals develop lung cancer by forcing them to smoke, and they have done the same by painting cigarette tar on mice. Does this prove the smoking theory? Not necessarily. The defence will say that the doses are unrealistically high. I have some sympathy with that view, but it certainly does not disprove the theory.
9. Analogy: "In some circumstances it would be fair to judge by analogy." I’ve never quite understood why this should be a scientific criterion so I’m going to ignore it, but if the defence wish to discuss it they should feel free.
This is my starting point. Crucially, why is the association so strong? If it is not smoking, then what? Why does the lung cancer epidemic follow in the footsteps of smoking in terms of location and gender? Or does it? Are we being sold a lemon? Over to you, the defence…