Here we go again. Yet another “study” that “suggests that among non-daily smokers, young adults who use e-cigarettes tend to smoke more cigarettes and to do so more frequently. Such individuals may be at greater risk for chronic tobacco use and dependence.”
Amusingly, this study was ‘accepted’ by the journal Preventative Medicine in March this year – which would have been around the time that the latest figures from the CDC was being compiled – that data was published in June.
That’s the thing with longitudinal survey studies, they are slow to gather data and even slower to get published. In this case, the data released by the CDC immediately throws the data from this study into doubt.
As with any longitudinal survey study, it relies heavily on participants actually completing the study. In this case, there was actually a fiscal incentive for participants to do just that.
Participants were recruited via paid online advertisements, primarily on Facebook. Advertisements appeared to users with profiles indicating they met age and residency criteria. Accompanying text indicated that eligibility criteria included recent smoking. Eligible and interested individuals provided informed consent, then completed the baseline assessment. Additional assessments were completed 3, 6, 9, and 12 months later. Participants were compensated $25 at baseline and 12 months, and up to $40 each at 3, 6, and 9 months.
Now, $25 isn’t a large amount of cash and neither is $40 (approximately £20 and £30 at today’s exchange rate). What I found interesting in this statement was the paid online adverts, primarily on Facebook.
As we know, the film A Billion Lives struggled to promote their film on Facebook as it was considered a “tobacco product”, so I have to wonder just how these advertisements were worded to avoid being blocked by the social media network. Of course, the tin-foil hat version is that it’s because it was investigating tobacco use, you know, for science.
So paying Facebook for targeted advertising space gave this lot 391 individuals to survey. Not exactly a large sample group, and definitely not a sample that can be extrapolated to population level. Nevertheless, the researchers make the usual wild claims about e-cigarette users (young adults) being more likely to smoke combustible cigarettes, and of course dual users smoking more than their peers who only use cigarettes.
Is any of that true?
Well, not really no.
Cigarette and e-cigarette use were assessed at screening, baseline, and 3, 6, 9, and 12 months post-baseline. At screening, participants were asked “How frequently have you used e-cigarettes in the past 6 months?” Response options included: 0 times; 1–3 times; 1–2 times per month; weekly; 2–4 times per week; and daily/almost daily (pre-baseline e-cigarette frequency). A comparable item was used to assess cigarette frequency over the previous 6 months (pre-baseline cigarette frequency).
So, unlike other US surveys and even UK surveys the measure of e-cigarette use appears to be arbitrary (which doesn’t surprise me considering the goal of the study – ‘The goal of this study was to test the hypothesis that use of e-cigarettes among young adult non-daily cigarette smokers would be associated with increased cigarette consumption.’ – is clear.)
Naturally, due to the number of variables the data on cigarette use/frequency and e-cigarette use/frequency was collapsed, albeit in startlingly different ways. Cigarette use variables included total cigarettes and cigarette days (quantity of cigs smoked and how often) at all 5 time points (baseline, 3, 6, 9 and 12 months). Conversely, e-cigarette data ended up being a binary variable (yes/no) which of course won’t reflect any level of detail such as: how frequently have you used an e-cigarette in the past x months/days – which is what they ask at baseline.
For example, a participant who endorsed e-cigarette use at every timepoint would have baseline, 3, 6, 9, and 12 month e-cigarette stability values of 1, 2, 3, 4, and 5, respectively. A participant who reported e-cigarette use only at 3 and 12 months would have values of 0, 1, 1, 1, and 2.We assume that, if e-cigarette use predicts increasing cigarette consumption over time, those who use e-cigarettes more consistently across multiple timepoints are more vulnerable to this effect.
So, if an individual uses an e-cigarette at 3 and 12 months (as per their example) they assume that at 6 & 9 months, that hasn’t changed – hence the values they cite as their example.
Thus, our analyses included e-cigarette stability as a predictor measuring aggregate e-cigarette use over time, rather than current or recent, but not cumulative, use. To account for differences in quantity and frequency of use due to variability between timepoints in the number of assessment days, we created a time-varying variable (assessment days) measuring the number of days on which use was assessed at each timepoint.
There, in black and white, is the problem. Instead of analysing current or recent use (as any sane researcher would do), they assume that use-once-use-always. Therefore, if an individual reported use at the 3 month interval, that use stat would carry over onto the following intervals.
To correct their example of an individual who used an e-cigarette at 3 & 12 months, the values would be 0, 1, 0, 0, 2. If they wanted to see actual use that is.
In terms of cigarette and ecigarette use over time, the proportions of data missing at 3 month, 6 month, 9 month, and 12 month timepoints were relatively low: 3%, 11%, 14%, and 9%, respectively. Eighty-seven participants (22%) had missing data at ≥1 assessments.
According to the paper, 5% (19) of the participants had missing data at 9 & 12 months, these were considered to have dropped out of the survey. Participants that stayed in for the duration but had missing data “did not differ from the rest of the sample in terms of demographics or pre-baseline or baseline cigarette or e-cigarette use“. The number that had missing data at any interval equated to 87 participants, and through assumption, these were analysed solely because they didn’t appear to differ from the rest of the sample.
Just a minor limitation then.
During the six months pre-baseline, 19% of participants reported no e-cigarette use, 32% 1–3 uses, 27% 1–2 uses/month, 10% weekly use, 6% 2–4 uses/week, and 6% daily/almost daily use.
So, 74 reported no e-cigarette use at all in the 6 months prior to the study commencement and only 23 reported regular use (daily/almost daily), while 316 (total) had reported any use.
During the 12-month study, 53% reported any e-cigarette use.
Wait, what? Oh, I forgot how they represented e-cigarette use.
Across timepoints, participants used e-cigarettes on 9–14% and cigarettes on 30–46% of days assessed.
So across all 5 time intervals, cigarette use remained much higher than e-cigarette use. Not surprising given that combustible tobacco is still the most popular in the US.
While cessation was not directly assessed, 44 participants (11.2%) denied smoking in the past 14 days at 12 months, and 23 of these (5.9%) had given the same response for the 9 days of assessment at 9 months.
Ah, here’s an interesting one. 44 participants (self-reported as smokers) denied smoking at the final interval, and just over half of those had given the same response at 9 months. That’s 44 ex-smokers.
Another interesting snippet to look at, considering California raised the legal age for purchasing “tobacco products” (including e-cigarettes, natch) to 21:
Neither cigarette nor e-cigarette use differed for those who were 18–20 when the law was enacted compared to older participants. Among the former, there were no changes in use after the restriction was implemented.
Which just goes to show how pithy and utterly pointless raising the purchasing age actually is.
The main effect indicates that each one-category increase in e-cigarette frequency (e.g., from 1 to 3 uses in 6 months to monthly use) predicted a 13% increase in total cigarettes at baseline (i.e., at time=0).
Now this is utterly confusing. How would an increase in e-cigarette usage at later time intervals (3, 6, 9 or 12 months) predict an increase in cigarette use at baseline?
At baseline, those who reported ≥4 e-cigarette uses in the previous 6 months were smoking 1.15 cigarettes per day, compared with 0.96 for less frequent e-cigarette users.
Nonsensical. But it gets worse.
We next examined the effect of pre-baseline e-cigarette use on cigarette days over time,with similar results.
So comparing pre-baseline e-cigarette use against total cigarette use (days) over the entire 12 months. Seems like a invalid observation. Unless of course you’re using it to emphasise a rather moot point:
more frequent pre-baseline use of e-cigarettes was associated with more frequent cigarette smoking over the next 12 months, but the effect size did not change over that period.
So if you used an e-cigarette more in the six months before the study commenced, you’re going to smoke more. Apparently. Even the “cigarette quantity” (according to these brainiacs) showed “the same” result:
across the year of follow-up, each additional time-point at which e-cigarette use was reported predicted 18% more cigarettes smoked. Comparing the opposite tails of e-cigarette stability, a participant who used e-cigarettes at every timepoint smoked more than twice as many cigarettes as a participant who never used e-cigarettes.
Thing is, the definition of “e-cigarette use” is too broad, in that it can be used to fit any narrative. According to this piece of ‘research’ (i.e. a self-selecting internet survey) where the majority of participants were current smokers and some dual-users (19% of the sample didn’t report e-cigarette use prior to the commencement of the study), more “stable” (strong and stable as opposed to weak and wobbly, natch) e-cigarette users would self-report an increase in cigarette smoking.
We tested pre-baseline and contemporaneous e-cigarette use as prospective predictors of two outcomes. As expected, those who used e-cigarettes more frequently over six pre-baseline months reported more cigarettes at baseline, and the gap grew larger over 12 months.
More frequent pre-baseline use of e-cigarettes also predicted greater frequency of cigarette days, but this effect did not change over time. Similarly, after accounting for pre-baseline cigarette and e-cigarette frequency, those who used e-cigarettes more consistently over 12 months also reported greater cigarette quantity and frequency.
In other words, they expected the outcome most likely due to how the questions were worded, and of course the measure of pre-baseline (and later on) e-cigarette “use”.
Contrary to hypotheses, e-cigarette stability was not consistently associated with 12-month growth in cigarette consumption. There are multiple mechanisms that could explain this inconsistency. One possibility is that young adults who use more cigarettes may find ecigarettes more attractive than lighter-smoking peers. That is, cigarette use may increase likelihood of e-cigarette use, rather than vice versa, although the use of propensity scores and low correlations (rs = 0.07–0.13) between e-cigarette and cigarette frequencies at individual timepoints make this less likely.
Of course, their number-crunching completely negates the possibility that heavier smoking youths are likely to find e-cigarettes more attractive. Well, that’s a bit of a no-brainer really, considering the prevalence of e-cigarettes among US youth (despite their best efforts to kick them off the market).
Further, the limitations of this paper aren’t limited to those mentioned in the paper itself:
Participants were young adults who had smoked cigarettes recently, limiting generalizability to other age groups or non-smokers. Ecigarette items did not explicitly define the product, or assess nicotine content. However, when we added an item at 12 months to address the latter, 90% of the 48 participants who completed it reported usually or always using nicotine products. Finally, the sample was recruited using Facebook advertisements targeted toward specific demographic characteristics. Participants were those who opted to learn more and to enroll; whether they are representative of non-daily smokers is unknown.
I’ll add the usual flaw of the measurement of “e-cigarette use”; only one survey in the US comes close to any reasonable form of measurement – the Monitoring the Future survey. Maybe, now that there is a recommended list of core items to measure & ask about, the results generated from surveys will be of any consequence.
Let’s put this study where it belongs eh?