Though not strictly trans-scientific, it turns out to be practically impossible to assess whether vaping is causing people to smoke (let alone become smokers) with any of the data that is apparently available. Even the best potential matching or deconfounder covariates are inadequate to control for propensity to use a tobacco product. The covariates used in practice are nowhere near close enough. Variables that are supposed to control for the propensity are typically measures of “risk taking” or “rebellious” behavior and perhaps inclination to use drugs of any sort. (These are often described with the phrase “common liability”, another term that should be avoided by ethical researchers because of the obvious pejorative double entendre of the word “liability”. Notice that in other contexts where there is a common cause -- the appropriate scientific term -- that makes someone more liable to choose both the exposure and outcome, researchers never use this loaded phrase.)
The variables used are terribly imprecise proxies for the overall propensities to take risks and use drugs. Moreover, the proxies should include the logistics of someone’s peer and adult social environment, which are at least as important as attitudinal inclinations, but do not. But even if “risk taking” proclivity were measured perfectly (which is far from the case) it could only explain such questions as why vaping is associated with motorcycle riding. It would still miss a major common cause of why vaping is associated with subsequent use of other drugs. There are inclinations, often traceable to specific psychological or environmental conditions but unlikely to be measured in a dataset, that cause a desire to pursue not “risk” or “rebellion” generally, but specifically pursue altered states of consciousness. This is the major, and effectively intractable, confounding problem for honest attempts to assess various drug gateway claims.
Yet it turns out that in the present case, the “risk taking” and “rebellious” variables plus wanting altered states are still not the most important nearly-intractable confounding problem. A “risk taker” might like motorcycles but not cannabis, someone seeking a particular altered state might like cannabis but not cocaine, and someone seeking altered states in general might still not like motorcycling. But almost everyone who likes the specific experience of smoking likes the specific experience of vaping because it is almost the same drug and almost the same behavior pattern. Researchers in this area, as well as most critics of this research, seem oblivious to this.
About half the population likes being under the influence of nicotine and half does not. This variation alone guarantees a substantially higher smoking uptake among vapers (and vice versa). This is partially a result of physiology and psychological characteristics, and partially a matter of attitude. Most people who do not use any nicotine product are actively averse to using any nicotine product. Thus people who never vape, smoke, nor use any other tobacco product will inevitably initiate one such product less often than users of one of the other products. This is no different from the fact that people who smoke one brand of cigarette are far more likely to subsequently smoke another brand than are people who do not smoke. No one would take that observation to mean that in a counterfactual world in which the first brand did not exist, the person never would have smoked any cigarette (i.e., the product that happened to be used first obviously did not cause someone to be inclined to use some product).
Attempts to control for this common cause are generally doomed. It is almost impossible to control for having a physical/psychological “taste” for nicotine, since the only good measures of it are already in use as the exposure and outcome variables. Smoking status of close relatives (getting benefits from nicotine seems to be heritable and attitudes toward it certainly are) is a very rough proxy for both tastes and attitudes, but clearly not good enough.
Much of the problem here seems to be a particularly bad result of the generic problem in tobacco research of not stating, much less testing, clear causal hypotheses. An expectation that a causal study must have a clear causal hypothesis would change the entire literature. It is difficult to figure out how to non-absurdly finish the sentence, “we hypothesize that many young people who find they like vaping then switch to the harmful, less flavorful, less socially acceptable, and more expensive (in all the jurisdictions where these studies happen) alternative -- something that they would not have otherwise ever done -- because....” The only apparent non-absurd conclusions are “...because their access to vaping products was interrupted while cigarettes remained easily available” or “...because someone convinced them that smoking was less harmful to their health.” It seems unlikely the worst papers would ever have been written if the authors had been required to finish that sentence.
The lessons available from the papers in this review are almost all negative. There are several papers that are solid workaday building blocks, but their generalizable lesson is basically just the negative “don’t overreach”. But most of the papers offer only errors to learn from. The questions most of the authors here are attempting to answer simply cannot be informed by the methods they chose. There are a couple of papers that are reasonably informative about some of their study questions. But we are sadly unable to point to a single example of someone using suitable methods to try to deal with the particular challenges of answering these questions.
.. There are, however, countless sources of good building-block information that can be pieced together to provide knowledge. It is thanks to such information that we know key facts, like vaping is approximately harmless and that it helps many people quit smoking. Part of the reason those salient facts are obscured by the journal literature is the political bias of the well-funded authors. But that is not the only reason. The typical bad practice in public health type research -- pretend your data is what you wish it were rather than what it is, thoughtlessly calculate simple associations, ignore all other information, and then declare sweeping conclusions as if the outputs from one logistic regression were divine revelation -- is what allows intentional propagandists to get away with junk science. But it also makes even honest efforts largely uninformative.
No comments:
Post a Comment