The authors have declared that no competing interests exist.
Conceived and designed the experiments: KF RG MQ. Analyzed the data: KF MQ. Wrote the paper: KF RG MQ.
An understanding of women’s longitudinal patterns of smoking during the pre-conception, pregnancy and postnatal period and the factors associated with these patterns could help better inform smoking cessation services and interventions.
Latent class analysis (LCA) was used to empirically identify women’s smoking patterns in a sample of 10,768 mothers from the 2010 UK Infant Feeding Survey. Multinomial logistic regression was used to identify characteristics associated with these patterns.
LCA identified five distinct smoking patterns during the pre-conception, pregnancy and postnatal period: “non-smokers” (74.1% of women); “pregnancy-inspired quitters” (10.2%); “persistent smokers” (10.1%); “temporary quitters” (4.4%); and postnatal quitters (1.1%). Smoking patterns varied markedly according to socio-demographic variables and parity. After adjusting for these variables, mothers who lived during pregnancy with a partner who smoked were more likely to be temporary quitters (aOR 2.64, 95% CI 1.74–3.99) or persistent smokers (aOR 3.32, 95% CI 2.34–4.72) than pregnancy-inspired quitters. Mothers who lived during pregnancy with someone else other than a partner who smoked were more likely to be persistent smokers (aOR 2.34, 95% CI 1.38–3.97) or postnatal quitters (aOR 2.97, 95% CI 1.07–8.24) than pregnancy-inspired quitters. Mothers given information on how their partner could stop smoking if they lived during pregnancy with a smoking partner were less likely to be persistent smokers (aOR 0.42, 95% CI 0.27–0.65) than pregnancy-inspired quitters.
Health professionals should ask about smoking at every opportunity, and refer women who self-report as current smokers to an evidence based smoking cessation service.
Smoking in pregnancy is associated with health risks for both the mother and infant, including increased risks of stillbirth, low birthweight and sudden infant death syndrome [
A range of effective interventions are available to help pregnant women stop smoking [
The aims of this study were therefore to identify women’s longitudinal patterns of smoking during the pre-conception, pregnancy and postnatal period and investigate what characteristics are associated with these patterns with a view to informing smoking cessation services and interventions.
The study sample comprised mothers who participated in the 2010 UK Infant Feeding Survey (IFS) [
At each stage of the survey, mothers were asked questions about their cigarette smoking behaviour: at stage 1, mothers who reported ever smoking cigarettes were asked to recall if they had smoked at all in the last two years, which roughly covers the period of their pregnancy and the year before conception. Women who responded “yes” to this question were asked “Do you smoke cigarettes at all now?” Those who responded “yes” were asked to recall if they smoked cigarettes at all during pregnancy, after they found out they were pregnant, whilst those who responded “no” were asked to recall when they finally gave up (ticking one of the following answers: before they knew they were pregnant; as soon as they found out they were pregnant; later on during the pregnancy; or after the birth). At stages 2 and 3, mothers were also asked the question “Do you smoke cigarettes at all now?” Information from these questions was used to identify mother’s smoking status (whether or not they smoked cigarettes at all) at six time points—one year before pregnancy, during pregnancy after confirmation of pregnancy, later in pregnancy and at stages 1–3 of the survey.
A range of socio-demographic and pregnancy-related variables was examined as potential variables that might be associated with smoking patterns based on the published literature [
Latent class analysis (LCA) is a statistical method which posits that homogenous unobserved subgroups (latent classes) can be identified within a heterogeneous group using a set of observed (indicator) variables. Using information on women’s observed smoking status at six time points (see six indicator variables in
Number (% |
|
---|---|
SMOKING STATUS OF MOTHER | |
Non-smoker | 8258 (73.5) |
Smoker | 2499 (26.5) |
Non-smoker | 9522 (87.7) |
Smoker | 1050 (12.3) |
Non-smoker | 9617 (88.7) |
Smoker | 955 (11.3) |
Non-smoker | 9625 (87.2) |
Smoker | 1130 (12.8) |
Non-smoker | 9547 (86.4) |
Smoker | 1221 (13.6) |
Non-smoker | 9435 (85.3) |
Smoker | 1333 (14.7) |
SOCIO-DEMOGRAPHIC CHARACTERISTICS OF MOTHER | |
35 or older | 2632 (19.4) |
30–34 | 3783 (28.5) |
25–29 | 2850 (28.0) |
Under 25 | 1477 (24.1) |
Married/civil partnership or cohabiting | 9538 (85.7) |
Single | 1059 (13.4) |
Widowed, divorced or separated | 87 (0.9) |
White | 9715 (86.6) |
Non-White | 815 (13.4) |
Over 18 | 6318 (51.9) |
17 or 18 | 2878 (30.1) |
16 and under | 1497 (18.0) |
Managerial & professional | 4696 (34.8) |
Intermediate | 2248 (19.6) |
Routine & manual | 2438 (26.7) |
Never worked | 567 (9.8) |
Not classified | 819 (9.1) |
PREGNANCY-RELATED CHARACTERISTICS | |
One child | 5307 (52.5) |
Two or more children | 5461 (47.5) |
Did not drink | 6204 (63.5) |
Drank less than one unit | 3112 (29.6) |
Drank one or more units | 658 (6.9) |
No | 8589 (77.5) |
Yes | 2179 (22.5) |
No | 10339 (93.4) |
Yes | 429 (6.6) |
No | 2703 (25.2) |
Yes | 7997 (74.8) |
No | 1275 (12.3) |
Yes | 895 (10.2) |
NA, did not live with partner who smoked | 8589 (77.5) |
b women in Northern Ireland were not asked their ethnic group since 99% of mothers in the 2001 census were white, therefore they are all assumed to be white.
The number of latent classes was decided by fitting models with different numbers of latent classes and then considering model interpretability and model fit, parsimony and stability using the Akaike information criterion (AIC)[
Number of latent classes | AIC | BIC | CAIC | adjBIC | Entropy R2 | G2 |
---|---|---|---|---|---|---|
1 | 29553.99 | 29597.7 | 29603.7 | 29578.63 | 1 | 29541.99 |
2 | 3077.222 | 3171.919 | 3184.919 | 3130.606 | 0.98402839 | 3051.222 |
3 | 1056.327 | 1202.014 | 1222.014 | 1138.456 | 0.97074606 | 1016.327 |
4 | 242.0042 | 438.6812 | 465.6812 | 352.8787 | 0.98395689 | 188.0042 |
5 | ||||||
6 | 183.5062 | 482.1639 | 523.1639 | 351.8713 | 0.70815914 | 101.5062 |
Multinomial logistic regression was used to investigate what variables were associated with latent class membership (i.e. smoking pattern as a categorical outcome variable). When examining the association between the socio-demographic/pregnancy-related variables and latent class membership, the models included all women–smokers and non-smokers. This was a statistically efficient way of simultaneously comparing non-smokers, women who successfully quit smoking after becoming pregnant and those who followed other smoking trajectories. A full regression model was developed by including socio-demographic and pregnancy-related variables that have previously been associated with women’s smoking behavior. Variables remained in the full model if there was evidence (p<0.05) that they were associated with membership in at least one latent class; these variables are listed in
Temporary quitters during preg (Class 2) vs Pregnancy-inspired quitters | Non-smokers (class 3) vs Pregnancy-inspired quitters | Persistent smokers (class 4) vs Pregnancy-inspired quitters | Postnatal quitters (class 5) vs Pregnancy-inspired quitters | |
---|---|---|---|---|
aOR |
aOR |
aOR |
aOR |
|
SOCIO-DEMOGRAPHIC VARIABLES | ||||
35 or older | 1 | 1 | 1 | 1 |
30–34 | 1.17 (0.70–1.98) | 0.94 (0.73–1.22) | 1.02 (0.66–1.58) | 1.3 (0.51–3.30) |
25–29 | 0.76 (0.58–1.01) | 1.31 (0.85–2.01) | 1.38 (0.53–3.60) | |
Under 25 | 1.77 (0.93–3.35) | 1.32 (0.81–2.14) | 0.75 (0.24–2.37) | |
Married/civil partnership or cohabiting | 1 | 1 | 1 | 1 |
Single | 1.55 (0.59–4.07) | |||
Widowed, divorced or separated | 0.46 (0.12–1.84) | 1 (0.33–3.02) | 0.67 (0.07–6.42) | |
White | 1 | 1 | 1 | 1 |
Non-White | 0.69 (0.31–1.57) | 0.75 (0.16–3.51) | ||
Over 18 | 1 | 1 | 1 | 1 |
17 or 18 | 0.87 (0.58–1.31) | 1.37 (0.94–2.00) | 1.03 (0.56–1.92) | |
16 and under | 1.07 (0.67–1.71) | 0.97 (0.42–2.28) | ||
Managerial & professional | 1 | 1 | 1 | 1 |
Intermediate | 0.64 (0.39–1.04) | 0.89 (0.57–1.40) | 0.62 (0.22–1.78) | |
Routine & manual | 1.1 (0.70–1.71) | 1.58 (0.58–4.34) | ||
Never worked | 1.73 (0.64–4.69) | 1.84 (0.97–3.51) | 2.28 (0.43–12.16) | |
Not classified | 0.9 (0.45–1.83) | 0.76 (0.51–1.12) | 1.59 (0.89–2.85) | 2.26 (0.74–6.93) |
PREGNANCY-RELATED VARIABLES | ||||
One child | 1 | 1 | 1 | 1 |
Two or more children | 0.97 (0.48–1.93) | |||
Did not drink | 1 | 1 | 1 | 1 |
Drank less than one unit | 1.27 (0.85–1.88) | 1.28 (0.92–1.78) | 1.46 (0.73–2.91) | |
Drank one or more units | 1.23 (0.68–2.22) | 1.06 (0.60–1.88) | 1.62 (0.64–4.11) | |
No | 1 | 1 | 1 | 1 |
Yes | 1.98 (0.96–4.10) | |||
No | 1 | 1 | 1 | 1 |
Yes | 1.17 (0.55–2.46) | 0.85 (0.51–1.41) | ||
No | 1 | 1 | 1 | 1 |
Yes | 0.91 (0.55–1.50) | 1.14 (0.52–2.54) | ||
No | 1 | 1 | 1 | 1 |
Yes | 0.71 (0.42–1.21) | 1.21 (0.84–1.73) | 0.67 (0.26–1.70) | |
NA, did not live with partner who smoked |
a Adjusted for all variables in the table.
b women in Northern Ireland were not asked their ethnic group since 99% of mothers in the 2001 census were white, therefore they are all assumed to be white.
* p<0.05
** p<0.01
*** p<0.001.
Our study was secondary data analysis of datasets from the IFS which are deposited in the UK Data Archive. The original survey was conducted by IFF Research and the University of York, on behalf of the government health departments of England, Northern Ireland, Scotland and Wales. Ethical approval to conduct the original survey was granted by the Ethics Committee, Department of Health Sciences at the University of York.
Using the information on women’s observed smoking status outlined in
Hence, an estimated 11.2% of mothers smoked during pregnancy (adding classes 4 and 5), and of the 1,579 mothers who quit during pregnancy (classes 1 and 2), 479 women (30.3%) relapsed by 8–10 months postnatally.
From a smoking intervention point of view it is of particular interest to know what distinguishes between women who successfully quit smoking after becoming pregnant and those who follow other smoking trajectories. Pregnancy-inspired quitters (n = 1,100) were therefore used as the reference group when examining the association between selected variables and latent class membership (i.e. smoking pattern) (
In unadjusted analysis (data not shown), all socio-demographic and pregnancy-related variables were significantly associated with membership in at least one of the latent classes. For example, about half of women who were persistent smokers, temporary quitters or postnatal quitters lived with a partner who smoked, compared with 37.8% of pregnancy-inspired quitters and 13.5% of non-smokers (
Compared to pregnancy-inspired quitters, persistent smokers were more likely to be single, have white ethnicity, have finished full-time education aged 16 or under, be in routine or manual occupations or never worked, and be parous (
By contrast, non-smokers were less likely than pregnancy-inspired quitters to be aged under 25yrs, single, widowed, divorced or separated (rather than married or co-habiting), have white ethnicity, have finished full-time education at a young age, be in intermediate or routine and manual occupations, be non-parous, and to have drank alcohol during pregnancy. After adjustment for these variables, they were less likely to have lived with a partner who smoked (aOR 0.23, 95% CI 0.18–0.30) and to have been given info about smoking during pregnancy (aOR 0.53, 95% CI 0.41–0.69).
Our study suggests that women exhibit one of five distinct patterns of smoking during the pre-conception, pregnancy and postnatal period: most prevalent (74.1%) were the “non-smokers”; an estimated 10.2% were “pregnancy-inspired quitters” with a tendency not to relapse postnatally; 10.1% were estimated to be “persistent smokers”; 4.4% were estimated to be “temporary quitters” with a tendency to relapse postnatally; and 1.1% were estimated to be “postnatal quitters”. Smoking patterns varied significantly according to socio-demographic variables and parity. After adjustment for these variables, we found that mothers who lived during pregnancy with a partner who smoked were more likely to be temporary quitters or persistent smokers than pregnancy-inspired quitters. Mothers who lived during pregnancy with someone else other than a partner who smoked were also more likely to be persistent smokers or postnatal quitters than pregnancy-inspired quitters. In addition, compared to pregnancy-inspired quitters, persistent smokers were less likely to have been given information on how their partner could stop smoking if they lived during pregnancy with a smoking partner.
Mumford et al [
The prevalence of smoking in pregnancy observed in our study is consistent with other UK data from the same period. For example, the proportion of mothers smoking later on during pregnancy in our study (11%,
Another large UK cohort of mothers who gave birth in 2000–2001 (22) observed a higher prevalence of relapse at 9 months postnatally (57%) than our estimate of 30% at 8–10 months postnatally. Similarly, US data from 10 states in the 1990s (23) suggests that half of women relapse by 6 months postnatally. The British study which used LCA to identify smoking patterns [
The characteristics our study found to be associated with women’s smoking patterns are consistent with previous reviews which suggest that predictors of smoking among pregnant women include younger maternal age [
A recent synthesis of 38 qualitative studies [
Key strengths of our study are that our findings are based on a large, contemporary, national cohort of mothers. We had a large enough sample size (n = 10,768) to identify five distinct smoking patterns and identify characteristics associated with them. Although the initial response rate was 51%, with some attrition at later stages, we used survey weights, which were based on a large number of variables, to take account of differential sampling, differential response rates among different groups and non-response bias introduced through attrition over the course of the survey. A potential limitation is that mother’s smoking status was based on self-report, which could have led to under-reporting given the social stigma of smoking in pregnancy [
Some women appear to exhibit marked fluctuations in smoking during the pre-conception, pregnancy and postnatal period suggesting the need for health professionals to ask about smoking at every opportunity, and refer women who self-report as current smokers to an evidence based smoking cessation service. Women at high risk of relapsing postnatally, such as those who are single or parous, can be identified antenatally as needing more support in the postnatal period. Our findings also suggest that pregnant women may be more likely to successfully quit smoking if partners and other household members who smoke are supported to stop smoking. Nearly a quarter of women (22.5%) lived during pregnancy with a partner who smoked; 6.6% of all women (39% of women aged under 20) lived with another household member who smoked). NICE [
(DOCX)
The authors would like to acknowledge Dr Laura Oakley, who provided input in the early stages of the study.