Participants
Pregnant women were recruited via 13 hospital antenatal clinics across the UK. All the hospitals were in localities that included areas of deprivation and social housing. The inclusion criteria for the trial were a desire to stop smoking, wanting help to stop smoking, agreeing to set a date for quitting smoking within one week of the baseline assessment, aged 16-50 years, being between 10 and 24 weeks of gestation, cigarette consumption of five or more daily before pregnancy, currently smoking one or more cigarettes per day, and being able to walk continuously for at least 15 min. Women were not eligible if they had health conditions that had potential to be exacerbated by exercise or advised not to exercise by a medical doctor, unable to provide informed consent or complete the study questionnaires in English, dependence on drugs or alcohol, and currently using or wanting to use nicotine replacement therapy. Written informed consent was obtained from all participants.
Design and interventions
This study reports analysis of depression outcomes from a RCT evaluating the effect of a physical activity intervention on smoking cessation during pregnancy [
21]. At the first session, participants were randomly assigned (based on a computer-generated code) to either behavioural cessation support alone (usual care) or to this support plus a physical activity intervention. In the primary analysis, the rates of continuous biochemically validated smoking cessation at end of pregnancy were similar in the physical activity group (7.7%) and usual care group (6.4%) [
24].
Intervention
Participants were offered the opportunity to take part in six weekly sessions of 20 min of individual behavioural support for smoking cessation, starting one week before the smoking quit date and ending four weeks later. This intervention seeked to support smoking cessation by reinforcing commitment to abstinence and solving participants’ problems about maintaining smoking abstinence. The intervention also aimed to improve the mental health of women.
The physical activity intervention combined supervised exercise sessions with physical activity consultations. All sessions were individual and took place in a private room at the hospital or in a community health centre or children’s centre. Fourteen sessions of supervised exercise were offered over eight weeks; twice a week for six weeks, followed by weekly sessions for two weeks. At each session participants were asked to walk at a moderate intensity on a treadmill for up to 30 min. At the first two treadmill sessions, and then on alternate occasions (total of nine consultations), they received the physical activity consultations, which aimed to identify opportunities to incorporate physical activity into women’s daily lives and to help them use behavioural strategies to improve adherence to these plans. These 20 min consultations incorporated 19 behaviour change techniques, as previously described in the protocol [
21]. Participants were advised to be active for at least 10 bouts, progressing towards 30 min of activity on at least five days a week, with an emphasis on brisk walking. As a motivational tool (not a research measure), participants were given a pedometer (Digi-Walker SW-200; Yamax, Nottingham, UK) and were encouraged to record the number of steps they had achieved each day, with the researcher calculating a 10% increment every two weeks; the overall goal being to work towards accumulating 10,000 steps per day [
27]. Participants received £7 for their travel expenses for each session that they attended and were given a DVD on antenatal exercise. On the other occasion the women received behavioural support for smoking sessions (up to six sessions) as for the usual care group.
Measures
One week prior to the quit date, baseline data was collected for demographic variables, smoking characteristics and physical activity behaviours, including the Fagerström Test for Cigarette Dependence score (FTCD, [
28,
29]) and self-reported levels of moderate and vigorous intensity physical activity (MVPA) in the previous week using a seven-day physical activity interview [
30]. Self-reports of physical activity included all types of activity, irrespective of context (e.g., leisure, occupational). Depression was measured with the Edinburgh Postnatal Depression Scale (EPDS [
31], via a face-to-face consultation at baseline and at end of pregnancy and via telephone at six months after the birth. The EPDS is a self-report 10-item scale (each is scored 0-3, range = 0 to 30) designed to assess antenatal and postnatal depression in community samples [
32], is widely used and has been validated in many countries [
33]. It can be used as a continuous outcome or to classify women as probable cases of antenatal/postnatal depression at a cut-off of 13 or above [
34]. Further self-reports of physical activity levels were collected at weeks one, four, and six after the quit date, at end of pregnancy and six months after the birth. In an 11.5% random subsample of participants (the target was 10%) physical activity was objectively measured using an accelerometer (Model GT1M or GT3X; Actigraph, Pensacola, FL, USA). This was worn over the right hip, in the fourth week after the quit date, for seven consecutive days, recording ‘dry land’ activity during waking hours at one minute epochs.
Analysis & Sample Size
Details of the sample size calculation can be found in previous publications. [
21,
24] First, we checked whether those providing EPDS data at the two follow-ups (end-of-pregnancy, six months postnatal) had similar baseline characteristics and physical activity at follow-up as the whole sample. Then we examined whether the baseline characteristics of the physical activity versus usual care group were similar in the sub-samples with EPDS data at the two follow-ups.
For the primary analysis EPDS data was treated as a continuous variable. We used a mixed-effect linear model with EPDS scores at end-of-pregnancy and six months postnatally as dependent variables. To estimate the difference between physical activity and usual care groups at each follow-up adjusting for baseline EPDS and recruitment centre, we fitted a linear mixed effect model including as independent variables the interaction of treatment groups and follow-up times, follow-up times, recruitment centre and the interaction of baseline EPDS score with follow-up times. The model accounts for within-person correlation over time and assumes that data is missing completely at random for the participants with missing EPDS data at follow-ups. In the next step, we further adjusted for the following potential predictors of postnatal depression: marital status, age at leaving full time education (as a proxy for socioeconomic status), body mass index and young age (i.e., age < 20 years). At end-of-pregnancy about half of the women (201/383, 52.5%) provided EPDS data before the birth and half after the birth (182/383, 47.5%); therefore we examined the difference between the mean EPDS scores before and after the birth using t-tests.
As a secondary analysis EPDS was treated as a binary variable. We assessed whether the proportion reaching the EPDS cut-off of ≥ 13 for depression changed at the follow-ups, for physical activity versus usual care, relative to baseline. We used a mixed effect logistic regression model, with the same independent variables as before, to estimate the adjusted odds ratios (OR) of depression for physical activity group versus usual at end-of-pregnancy and 6 month follow-up. This analysis was further adjusted for the same predictors used in the analyses of EPDS as a continuous variable. One recommendation is to use a EPDS cut-off of ≥ 15 antenatally and ≥ 13 postnatally [
35], therefore we conducted a further analysis using ≥ 15 for baseline and end-of-pregnancy follow-up before the birth, and ≥ 13 for end-of-pregnancy after the birth. Additionally, in order to explore whether adverse events might have influenced the EPDS outcomes we report these events for the two study groups.
We used multiple imputation analysis to explore the impact of missing data Missing values in the EPDS score at follow-up were replaced by imputed values using chained equations [
36] with the predictive mean matching method on the basis of the baseline explanatory variables of EPDS scores at baseline, randomisation groups, body mass index, carbon monoxide level, smoking abstinence at end-of-pregnancy, self-reported minutes of physical activity at baseline, age at leaving full-time education, gestational age (weeks) at baseline, number of cigarettes smoked before pregnancy, number of cigarettes smoked at baseline, FTCD score, marital status, ethnicity, type of physical activity at baseline, parity, partner smoking status, self-efficacy, confidence for physical activity, recruitment centre, perceived positive effects of being physically active. We created twenty imputed datasets and used the same model as above to estimate the treatment effects on EPDS score in these datasets. We combined the imputation-specific estimates using Rubin’s rules [
37]. All statistical analyses were performed using Stata (version 12).
To explore whether the effect of treatment on depression scores differed according to how physically active the women were at baseline, we tested for an interaction between baseline physical activity (< 150 min/week MVPA versus ≥ 150 min/week MVPA) and the treatment effect for the primary outcome of EPDS score at both follow-ups.
We also explored whether those who adhered better to the physical activity regime had a greater response to treatment by looking at modification of the treatment effect according to whether the individual reported ≥ 150 min/week MVPA at four weeks or six weeks after the quit day in those reporting < 150 min/week MVPA at baseline.