Characterising smoking and nicotine use behaviours among women of reproductive age: a 10-year population study in England | BMC Medicine

Pre-registration

The study protocol and analysis plan were pre-registered on Open Science Framework (https://osf.io/em8g2). We made one amendment prior to peer review. We had planned to analyse time trends with survey month modelled using restricted cubic splines with five knots. However, for analyses of trends in current use of nicotine pouches, we reduced this to three knots to avoid overfitting, because pouch use was only assessed over a relatively short period (November 2020–October 2023) and prevalence was assumed to be zero before this, based on previous evidence [19].

Design

Data were drawn from the Smoking Toolkit Study, an ongoing monthly cross-sectional survey of a nationally representative representative sample of adults (≥ 16 years) in England [20]. The study uses a hybrid of random probability and simple quota sampling to select a new sample of approximately 1700 adults each month. Interviews are held with one household member in selected geographic output areas until quotas are fulfilled. The quotas are based on factors influencing the probability of being at home (i.e. working status, age and gender). This hybrid form of random probability and quota sampling is considered superior to conventional quota sampling. Here, the choice of households to approach is limited by the random allocation of small output areas and rather than being sent to specific households in advance, interviewers can choose which households within these small geographic areas are most likely to fulfil their quotas. Therefore, unlike random probability sampling, it is not appropriate to record the response rate in the Smoking Toolkit Study.

Data were collected monthly through face-to-face computer-assisted interviews up to February 2020. However, social distancing restrictions under the COVID-19 pandemic meant that no data were collected in March 2020, and data from April 2020 onwards have been collected via telephone. The telephone-based data collection relies upon the same combination of random location and quota sampling, and weighting approach as the face-to-face interviews and comparisons of the two data collection modalities indicate good comparability [21,22,23].

For the present study, we used data from respondents to the monthly survey over a 10-year period from October 2013 to October 2023 (the most recent data available at the time of analysis). We restricted the sample to those aged ≥ 18 years as 16 and 17-year-olds were not surveyed in all waves. Our primary focus was women of reproductive age (which we defined as per the Office for National Statistics [1] as up to 45 years). We also provided data on these outcomes among the entire adult population in England for context.

Measures

Smoking status

Participants were asked which of the following best applies to them:

1. a)

   I smoke cigarettes (including hand-rolled) every day

2. b)

   I smoke cigarettes (including hand-rolled), but not every day

3. c)

   I do not smoke cigarettes at all, but I do smoke tobacco of some kind (e.g. pipe, cigar or shisha)

4. d)

   I have stopped smoking completely in the last year

5. e)

   I stopped smoking completely more than a year ago

6. f)

   I have never been a smoker (i.e. smoked for a year or more)

For analyses of current smoking, those who responded a-c were considered current smokers (coded 1) and those who responded d–f non-smokers (coded 0). For (unplanned) analyses of non-daily smoking (see statistical analysis section), those who responded b were considered non-daily smokers (coded 1) and all others (i.e. daily smokers or non-smokers) were coded 0. For analyses of quit attempts, those who responded a-d were considered past-year smokers and those who responded e-f were excluded.

Use of non-combustible nicotine products

Several questions asked participants about use of a range of nicotine products. Current smokers were asked ‘Do you regularly use any of the following in situations when you are not allowed to smoke?’; past-year smokers were asked ‘Can I check, are you using any of the following either to help you stop smoking, to help you cut down or for any other reason at all?’; and non-smokers were asked ‘Can I check, are you using any of the following?’. Those who reported using e-cigarettes in response to any of these questions were considered current vapers; those who reported using NRT (nicotine gum, lozenges/tablets, inhaler, nasal spray, patch, or mouth spray) current NRT users; those who reported using HTPs (‘heat-not-burn cigarette (e.g. iQOS, heatsticks)’) current HTP users; and those who reported using nicotine pouches (‘tobacco-free nicotine pouch/pod or ‘white pouches’ that you place on your gum’) current nicotine pouch users.

HTPs were included in the list of response options from December 2016 and nicotine pouches from November 2020; given the low prevalence of use of these products [19, 24], we imputed missing values as 0 (indicating no use) for participants surveyed before the response options were introduced. As a sensitivity check, we reran these models from the time when these data were available (i.e. December 2016 onwards for HTPs and November 2020 onwards for pouches); the results were unchanged.

Main type of cigarettes smoked

Current smokers were asked ‘How many cigarettes per day do you usually smoke?’ and ‘How many of these do you think are hand-rolled?’. Main type of cigarettes smoked was defined as hand-rolled for those reporting at least 50% of their total cigarette consumption is hand-rolled, and manufactured for those reporting that less than 50% is hand-rolled. This definition has been used in previous studies [25,26,27] and allows inclusion of those who smoke both hand-rolled and manufactured cigarettes.

Level of cigarette dependence

Current smokers were asked to self-report ratings of the strength of urges to smoke over the past 24 h [not at all (coded 0), slight (1), moderate (2), strong (3), very strong (4) and extremely strong (5)]. This variable was also coded ‘0’ for smokers who responded ‘not at all’ to the (separate) question: ‘How much of the time have you spent with the urge to smoke?’ [28]. This measure has been validated and performs at least as well as the Fagerström Test of Cigarette Dependence and the Heaviness of Smoking Index in predicting smoking cessation while not being subject to bias due to population-level changes in cigarette consumption over the time period of the study [28]. Scores were skewed towards lower values so we log-transformed this variable for analysis (with values of 0 imputed as 0.01 before the transformation was applied) and reported results as geometric means.

Quit attempts

Past-year smokers were asked: ‘How many serious attempts to stop smoking have you made in the last 12 months? By serious attempt I mean you decided that you would try to make sure you never smoked again. Please include any attempt that you are currently making and please include any successful attempt made within the last year’. Those who reported making at least one serious quit attempt in the past year were coded 1, else they were coded 0.

Success of quit attempts

Past-year smokers who had made an attempt to quit in the past year were asked: ‘How long did your most recent serious quit attempt last before you went back to smoking?’ Those who reported that they were still not smoking were coded 1, else they were coded 0.

Occupational social grade

Occupational social grade was defined according to the National Readership Survey classification [29] and categorised as ABC1 (includes managerial, professional, and upper supervisory occupations) and C2DE (includes manual routine, semi-routine, lower supervisory, and long-term unemployed). This occupational measure of social grade is a valid index of SES, widely used in research in UK populations, which is particularly relevant in the context of tobacco use [30].

Statistical analysis

Data were analysed in R version 4.2.1. Participants with missing data on key variables were excluded on a per-analysis basis (see Table 1 footnote for details). The Smoking Toolkit Study uses raking to weight the sample to match the population of England in terms of key demographics. These key demographics are determined each month using data from the UK Census, the Office for National Statistics mid-year estimates, and the National Readership Survey [20]. The following analyses used weighted data.

Where there were sufficient data, we used regression models (logistic/linear as appropriate, using the ‘svyglm’ command) to estimate monthly time trends in each outcome among women of reproductive age, overall and by occupational social grade. For the overall analysis, models only included time as an independent variable. For the analysis by occupational social grade, models included time, social grade, and their interaction as independent variables — thus allowing for time trends to differ across social grades. Time (survey wave) was coded 1…n where n was the total number of months in the time series (including March 2020 when no data were collected). Time was modelled continuously using restricted cubic splines with five knots (placed at equal quantiles of the data), to allow relationships with time to be flexible and non-linear, while avoiding categorisation. We were unable to model the interaction between time and occupational social grade for use of HTPs and nicotine pouches because very few women of reproductive age in the sample reported using these products at this time. We repeated these models using data from all adults (≥ 18 years) in England, to provide context.

We used predicted estimates from our models to (i) plot the prevalence (or geometric mean, for level of cigarette dependence) of each outcome over the study period (overall and by social grade, among women of reproductive age and in the entire adult population), and (ii) derive up-to-date estimates of the prevalence of each outcome in October 2023. We followed the ‘New Statistics’ approach to reporting and interpretation of results [31, 32], focusing on effect sizes and confidence intervals rather than dichotomous thinking about statistical significance (i.e. whether a result is significant or not significant, based on an arbitrary threshold). Where confidence intervals overlap, we report changes as ‘uncertain’.

In addition to our pre-registered analyses, where there was evidence that the trend in an outcome among women of reproductive age differed from the trend in the entire adult population, we repeated the model among men of the same age (18–45 years). This allowed us to explore whether the difference in trends was due to age more generally or was specific to women of reproductive age. We also added two unplanned analyses following peer review. In the first, we modelled time trends in non-daily smoking, to explore whether changes in current smoking we observed may have been driven by changes in non-daily smoking specifically. In the second, we modelled time trends in dual use of tobacco and non-combustible nicotine (i.e. current smoking and current use of e-cigarettes, NRT, HTPs, or nicotine pouches) as an additional outcome.