Discover more from Overcoming Bias
Random Smoking Trials
Hal Finney recently commented:
[Johnstone & Finch’s] Scientific Scandal of Antismoking … makes the case that smoking is not bad for your health. … [It has] the superficial appearance of referencing scientific studies and claiming the the mainstream misrepresents the results.
Yes, they are superficially credible. Their New Scientist letter:
WHO … claims … “an epidemic of chronic illnesses … could be prevented through simple changes in diet, by being more active and by not smoking.” … There have been a number of such studies, with various combinations of these three lifestyle factors, including the WHO collaborative trial (60,881 subjects, 6 years), the Goteborg trial (30,022 subjects, 11.8 years) and the Multiple Risk Factor Intervention trial (12,866 subjects, 7 years). These and another eight trials were conducted over three decades, one of the most expensive and sustained series of biological experiments in the history of medical science. … None showed any improvement in life expectancy and two showed a significant reduction in life expectancy in the test group.
So I dug further; bottom line: Johnstone & Finch are right. We usually see strong correlations between death and smoking, and we see those same correlations within each random arm (i.e., group) of a randomized trial. Nevertheless, we see no significant net death differences between control arms and arms induced to smoke less.
So we don’t have clear evidence that smoking kills on net; it could be that most or all of the death-smoking correlation is due to selection effects, and not smoking causing death. Experts say there is a substantial causal component, and for now I’m accepting that claim, but this lack of clear evidence is suspicious, and disturbing. Now for some details.
First I know of only one randomized trial directly and only on smoking. Of 1400 men, a random half were advised to stop smoking, and stop they did:
A randomised controlled trial of anti-smoking advice in 1445 male smokers, aged 40-59, at high risk of cardiorespiratory disease. After one year reported cigarette consumption in the intervention group (714 men) was one-quarter that of the “normal care” group (731 men); over 10 years the net reported reduction averaged 53%. The intervention group experienced less nasal obstruction, cough, dyspnoea, and loss of ventilatory function.
During the next 20 years there were 620 deaths (231 from coronary heart disease), 96 cases of lung cancer, and 159 other cancers. Comparing the intervention with the normal care group, total mortality was 7% lower, fatal coronary heart disease was 13% lower, and lung cancer (deaths+registrations) was 11% lower.
However, these differences are clearly not statistically significant. (The death difference 320-300 = 20 is a typical deviation sqrt(320) = 18.)
All the other smoking trials are randomized comparisons of whole packages of lifestyle advice; smoking is only one part of the package. The most recent such trial was MRFIT;
The MRFIT design called for the recruitment of at least 12,000 men aged 35 to 57 years who were at increased risk of death from CHD [heart attack], but had no clinical evidence of CHD. Persons were designated at “increased risk” if their levels of three risk factors-cigarette smoking, serum cholesterol, and blood pressure (BP)-were sufficiently high at a first screening visit to place them in the upper 15% of a risk score distribution based on data from the Framingham Heart Study. After about one third of the screening was completed and success in recruitment had been demonstrated, the 15% was changed to 10%.
Out of 8000 smokers, a random half were advised to quit smoking, and more of them did quit:
4,103 special intervention [SI] and 4,091 usual care [UC] men … reported smoking cigarettes at the first screening visit. Among SI men, the reported cessation rate increased from 43.1% at 12 months to 48.9% at 72 months. The reported cessation rate among UC men increased from 13.5% at 12 months to 28.8% at 72 months. Among smokers who reported cessation at 72 months, 51.3% of SI men and 22.7% of UC men had quit smoking within the first year and remained abstinent thereafter. Average thiocyanate and expired-air carbon monoxide served as objective measures of smoking and were significantly lower among the special intervention men.
Within each half of the experiment, the study did see strong correlations between death and smoking:
For both SI and UC men, substantial differences in subsequent CHD (34-49%) and all-cause (35-47%) mortality were evident for men who reported cigarette smoking cessation by the end of the trial compared with those continuing to smoke.
But comparing the two halves, we find that smoking less has no observed effect on lung cancer deaths:
After 16 years of follow-up, lung cancer mortality rates were higher in the SI than in the UC group … 135 SI and 117 UC participants died from lung cancer. … None of the hypotheses proposed to explain the unexpected higher rates of lung cancer mortality among SI as compared with UC subjects were sustained by the data.
Nor were there significant overall mortality effects from the entire package of advice:
After 16 years … 991 SI and 1050 UC men had died by the end of follow-up (relative difference, -5.7%; 95% CI, -13% to 2.8%)..
Again, those who smoked less died about 6% less (though more of lung cancer), but this difference still isn’t 5% significant.
In a randomized five-year multifactorial primary prevention trial of vascular diseases, hyperlipidemias, hypertension, smoking, obesity, and abnormal glucose tolerance of the high-risk test group (n=612 men) were treated with dietetic-hygienic measures and hypolipidemic (mainly probucol and clofibrate) and antihypertensive (mainly diuretics and β-blockers) agents. [There was] a matched high-risk control group (n=610) … Despite the highly significant reduction in the risk factor level, the five-year intervention program did not reduce coronary mortality or morbidity. In fact, the number of total coronary events tended to be higher in the intervention group than in the control group (19 vs nine cases; P=.057).
Bottom line: when folks freely choose if to smoke, we see strong correlations between death and smoking. The simple correlation is that smokers die about 60-70% more often, but after controlling for a few other factors this drops to 20-40%. Yet when a randomized trials selects 8000 of the 10% highest risk smokers, and induces half of them to smoke less, so that 49% instead of 29% quit smoking by seven years, that half dies only 6% less over 16 years, and dies more of lung cancer.
Now if that entire 6% drop were real and due entirely to one half smoking 20% less, that would translate into a roughly 30% smoking mortality effect. So if we assume high risk smokers aren’t hurt more by smoking than other smokers, we aren’t quite at the point of seeing a clear contradiction. Apparently we need bigger trials if we are going to see clearly if smoking kills on net. Alas the era of the large risk trial seems to be over, at least for now; it seems it will be a long time before we really know.