Purdue, Ohio State and Penn State faculty, led by Brian C.
Kelly, published an article in the April 2023 edition of Tobacco Control
titled, “E-cigarette
use among early adolescent tobacco cigarette smokers: testing the disruption
and entrenchment hypotheses in two longitudinal cohorts.” They concluded, “there is evidence
e-cigarette use among early adolescent smokers in the UK and USA leads to
higher odds of any smoking and more frequent tobacco cigarette use later in
adolescence.”
Lacking access to the U.K. data used by the authors, I
focused on the findings from the U.S., which were based on the federal
Population Assessment of Tobacco and Health (PATH) survey. To clarify details of the authors’ methods, I
reached out to them with some basic questions.
I telephoned and emailed the authors numerous times over seven
weeks from April to June, and copied my last email request to journal editor
Ruth Malone. She responded, suggesting
that I submit my questions via the journal’s Rapid Response system, which I
did on June 12. On June 30, the authors finally responded, with less than full
disclosure.
My questions are numbered, and they are followed by the authors’
responses in quotes, and then by my analysis in bold of what those responses
mean for the study’s results and conclusions.
1. Was public use or restricted
PATH data used? This is important, since Table 2 contains a cell, n=7, that is not
generally approved by NAHDAP.
“…the
analyses were based on the public use data from…the PATH Study.”
It
is critical to clearly identify the data used, which wasn’t the case in the
article.
2. Was the PATH cohort drawn from
Waves 1 and 4, with follow-ups to age 17 years as needed from the other waves?
“the
PATH study sample was drawn from the original cohort, the replenishment cohort,
and the shadow cohorts (see 1st and 2nd paragraphs of Methods Section).”
The
1st paragraph described the British cohort; it had nothing to do
with PATH. The 2nd paragraph
is a general description of the PATH youth survey, and while it mentions the
replenishment and shadow cohorts, it does not state that the authors actually
used them.
3. There were significant
differences in youth smoking-vaping between Wave 1 (2013-14) and Wave 4
(2016-18) that might have affected the results. Was each wave analyzed
separately as well as together?
The
authors didn’t answer this question.
4. The analysis included a
variable relating to “parent(s) smoking of cigarettes, cigars, or pipes.” Did
the analysis include other combustible tobacco product consumption by the
subjects themselves?
The authors failed to respond
to this question, or question #5 below.
5. Did the authors account for
age at first smoking or vaping (public use, < 12 years and 12-14 years) or
which product(s) had been used first?
The authors instead provided the
following statement:
“Regarding the remaining questions,
please note that our stated goal was to make the MCS [the British dataset] and
PATH analytical samples as comparable as possible when testing our hypotheses
using both cohorts (3rd paragraph of Methods section). As we note in the limitations section (5th
paragraph of Discussion section), the MCS had relatively limited items on
e-cigarette use and tobacco smoking compared to PATH. The MCS did not assess
other combustible tobacco product consumption in early adolescence, nor did MCS
measure the sequencing of early adolescent tobacco and e-cigarette use (noted
in the limitation section).”
The 3rd paragraph of the Methods section says that they
“restricted the PATH cohort to adolescents who were approximately the same age
as youth in the MCS cohort.”
While the authors noted the “limited items” and lack of
“sequencing” in the MCS data, they did not specify that they stripped the rich
PATH data of the comparable variables relating to #4 and #5 above. Since the PATH and British datasets were
entirely different, there was no scientific reason to strip PATH to make it
“comparable” to the MCS. It appears that
the authors took this action only to support their predetermined finding.
In summary, this was another questionable article claiming
that vaping leads to smoking. The
authors’ non-answers to my questions make it clear that they engineered this
study to produce their desired result.