The London Exercise And Pregnant smokers (LEAP) trial: a randomised controlled trial of physical activity for smoking cessation in pregnancy with an economic evaluation

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 07/01/14. The contractual start date was in February 2009. The draft report began editorial review in May 2014 and was accepted for publication in May 2015. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

The following coauthors have undertaken research or consultancy for companies that develop and manufacture smoking cessation medications: Robert West (Johnson & Johnson, GlaxoSmithKline, Pfizer), Tim Coleman (Pierre Fabre Laboratories) and Paul Aveyard (Pfizer).

Copyright statement

© Queen’s Printer and Controller of HMSO 2015. This work was produced by Ussher et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

The problem of smoking in pregnancy

Maternal smoking in pregnancy is the main preventable cause of morbidity and death among women and infants. Smoking is associated with adverse pregnancy and birth outcomes, including miscarriage, stillbirth, prematurity, low birthweight, congenital abnormalities, and neonatal or sudden infant death. 1–3 Smoking also presents immediate risks for the mother, including placental abruption,4 as well as the longer-term risks reported for smokers in general. In addition, the children of mothers who smoke are twice as likely to become smokers. 5 Smoking in pregnancy is a global public health problem. In high-income countries the prevalence of smoking in pregnancy is typically between 10% and 25% and it appears to be reducing. 6–10 However, rates seem to be rapidly increasing in low- and middle-income countries. 11 In the UK it is estimated that 12% of women smoke during pregnancy7 and, as in other high-income countries, rates of smoking in pregnancy remain highest among younger women and those who are more socially disadvantaged. 7 Smoking cessation during pregnancy improves maternal and birth outcomes,12 yet only approximately 25% of pregnant smokers stop for at least part of their pregnancy and around two-thirds of these relapse after giving birth. 13

Treatments to aid smoking cessation in pregnancy

Face-to-face and ‘self-help’ behavioural support are the only two interventions that have been shown to help pregnant women to stop smoking. 12,14 Regular sessions of face-to-face behavioural support can increase smoking cessation rates in pregnancy by approximately 6%15 and there is a need to identify other interventions that are effective during pregnancy when combined with this support. The most effective therapy in non-pregnant smokers is a combination of behavioural support plus nicotine replacement therapy (NRT), bupropion or varenicline. 16–18 However, the efficacy of NRT during pregnancy is not known,19 and thus many pregnant women are reluctant to use it,20 and other smoking cessation medications are contraindicated during pregnancy. 19 There is a need to identify other non-pharmacological interventions that are effective for smoking cessation during pregnancy.

Evidence for physical activity aiding smoking cessation

Effective pharmaceutical aids for quitting are thought to work mainly through reducing cigarette cravings18 and there is good evidence from a meta-analysis21 to show that physical activity (PA) reduces these cravings, particularly at a moderate or vigorous intensity. Therefore, PA interventions could aid smoking cessation. For non-pregnant smokers, a Cochrane systematic review22 has considered the evidence for PA aiding cessation. The majority of the 15 randomised controlled trials (RCTs) reviewed had low statistical power to detect a meaningful difference between the treatment groups, with seven trials having < 25 participants in each treatment arm. Six adequately powered trials compared a group receiving a PA intervention combined with behavioural support with a group receiving behavioural support alone. Three of these studies showed significantly higher smoking abstinence rates in the PA group than in the control group at the end of treatment. 23–25 One of these studies also showed that a PA intervention increased abstinence compared with a control group at the 3-month follow-up and there was a benefit of exercise of borderline significance [relative risk (RR) 2.19, 95% CI 0.97 to 4.96; p = 0.05] at the 12-month follow-up. 23 A further study showed significantly higher abstinence rates for the exercise group than for the control group at the 3-month follow-up, but not at the end of treatment or at the 12-month follow-up. 26 The study with the most intensive PA intervention, entailing thrice-weekly sessions of supervised vigorous-intensity exercise, showed the strongest effect on abstinence. 23 The other studies involved PA interventions that were relatively less intense, particularly in terms of the extent of supervised exercise, and it is possible that supervised exercise is needed for efficacy. Adequately powered trials are needed that involve moderate-intensity exercise, which is likely to be more acceptable than vigorous exercise for most individuals. 27 Moderate-intensity PA (e.g. brisk walking) is recommended for pregnancy28 and has been shown to reduce cigarette cravings during pregnancy,29 and pilot work suggests that pregnant smokers are likely to be receptive to a PA intervention. 30

The effects of physical activity on maternal depression and weight gain

Important secondary outcomes included changes in maternal depression and weight. Antenatal and postnatal depression are important because they are common and are associated with harmful consequences for the mother and child. 31–38 Interventions are needed for preventing and treating these types of depression. Moreover, pregnant smokers are at a heightened risk of depression during and after pregnancy, and women who quit smoking during pregnancy are more likely to relapse if they experience depressive symptoms. 12,39,40 Thus, it is important that pregnant women who smoke or who are attempting to quit are offered effective interventions for depression. The London Exercise And Pregnant smokers (LEAP) trial was the first study to assess the effectiveness of a PA intervention for symptoms of antenatal and postnatal depression specifically among smokers.

Excessive gestational weight gain (GWG) is associated with adverse pregnancy outcomes, including large-for-gestational-age infants and caesarean section. 41,42 In addition, smoking cessation is associated with GWG. 43 Interventions for managing GWG are needed, especially among women attempting to quit smoking, and PA has potential in this regard. Observational studies have demonstrated an association between participation in PA and reduced risk of excessive GWG,44–46 whereas a recent meta-analysis using data from 10 RCTs showed an overall benefit of participation in PA compared with a control condition in terms of reducing GWG. 47 We are not aware of any studies that have examined the effect of a PA intervention on GWG in pregnant smokers. Among non-pregnant smokers, there is some evidence that PA interventions can limit post-smoking cessation weight gain. 48 The LEAP trial was the first large RCT to examine the effect of a PA intervention on preventing excessive GWG and postnatal weight retention.

Summary

In summary, smoking in pregnancy is extremely harmful for mother and baby and is an enduring global public health problem. Behavioural support is the only smoking cessation intervention shown to be effective in pregnancy. The evidence for PA programmes aiding smoking cessation is mixed and pregnancy provides a compelling rationale for their use because medication is contraindicated or ineffective. We conducted the LEAP RCT to assess the effectiveness of a PA intervention for smoking cessation during pregnancy.

Main objective

The main objective of the study was to investigate whether or not standard behavioural support for smoking cessation in pregnancy plus a PA intervention is more effective than behavioural support alone in achieving biochemically validated smoking cessation between a quit date and the end of pregnancy for women between 10 and 24 weeks’ gestation who currently smoke one or more cigarettes daily and who smoked at least five cigarettes daily before pregnancy. A further objective was to assess the cost-effectiveness of the intervention for achieving smoking cessation at the end of pregnancy.

Trial design

The LEAP trial was a multicentre, pragmatic, randomised controlled, parallel-group trial of a PA intervention. Participants were monitored from their recruitment at between 10 and 24 weeks’ gestation until the end of pregnancy and were then followed up by telephone at 6 months after the birth. The trial protocol has been published. 49

Participants and recruitment

Interventions

The interventions followed Consolidated Standards of Reporting Trials (CONSORT) guidelines for non-pharmacological interventions. 53,54 Delivery of the interventions was standardised by training and by the therapists following manuals (see Appendices 2 and 3). The initial competence of the therapists was assessed by the trial manager/chief investigator by observing role-play scenarios during training. The fidelity of the interventions was monitored during the first 6 months by regular observations (at least five intervention sessions) by the trial manager/chief investigator, with the tasks listed in Appendices 2 and 3 used as a checklist. All sessions were face to face and one to one and were delivered in a private room at the hospital or in a community health centre/children’s centre. Social cognitive (learning) theory55 was the theoretical basis for the interventions. This theory recognises the interplay of individual factors (e.g. self-efficacy to quit smoking or increase PA) and social/environmental factors (e.g. social support) in health behaviour change. For each session that they attended, the women were paid £7 for their travel expenses.

Randomisation and blinding

An independent statistician generated a randomisation list using Stata version 11.2 (StataCorp LP, College Station, TX, USA), with random permuted blocks of random size stratified by recruitment centre. At enrolment the sequence was concealed from researchers, who had to confirm consent and eligibility on an online database before allocation was revealed. The online database was created by the Nottingham Clinical Trials Unit (NCTU) and held on a secure server in accordance with their standard operating procedures. Allocation was concealed from the participant until all baseline assessments had been completed. The sequence of treatment allocations was concealed until interventions had all been assigned and recruitment, data collection and laboratory analyses were complete. Thus, neither participants nor researchers were blinded to treatment allocation during intervention delivery or during outcome assessment.

Data collection

At baseline, researchers recorded demographic characteristics (including age, marital status, number of children, highest educational qualification, ethnicity, occupation, weeks of gestation and history of premature births) and smoking characteristics [including cigarettes smoked per day (now and before pregnancy), weekly urge to smoke (combining ratings of strength and frequency of urges),66,67 cigarette withdrawal symptoms,66,67 Fagerström Test for Cigarette Dependence (FTCD) score68,69 and partner’s smoking status]. Depression was assessed with the 10-item Edinburgh Postnatal Depression Scale (EPDS). 70 PA levels in the previous week (bouts of ≥ 10 minutes) were assessed for both groups using the 7-day PA interview (see Appendix 5). 71 Confidence levels with regard to taking up regular PA72 and stopping smoking73 were also recorded. Clothed weight (without shoes) was measured on a digital scale at the first antenatal booking visit by the midwife. The questionnaire showing assessments at baseline, including assessments repeated at further time points, is provided in Appendix 6. The timing of data collection is provided in Outcome measures.

Recording of adverse events

During all contacts, participants were asked about AEs. Medical records were examined monthly by research midwives for AEs and after delivery for maternal and infant outcome data. Researchers then summarised the descriptions in the case report forms and in the online study database. Descriptions were used to code the AEs according to standard terms in the Medical Dictionary for Regulatory Activities [see www.meddra.org (accessed 25 August 2015)]. Fetal deaths were recorded including miscarriage (non-live birth before 24 weeks’ gestation), stillbirth (non-live birth at ≥ 24 weeks’ gestation) and neonatal death (i.e. from live birth to 28 days).

Outcome measures

Statistical methods

The statistical analysis plan, which is presented in Appendix 7, was finalised before any analyses started. The analysis for the primary outcome was conducted by an independent statistician, with allocation to the two study groups concealed until the analysis was completed. Analyses were performed using Stata version 11.2 and IBM SPSS Statistics version 19 (IBM Corporation, Armonk, NY, USA). Throughout, a p-value of < 0.05 was taken to indicate statistical significance and 95% confidence intervals (CIs) were calculated.

Ethics and governance

An independent Trial Steering Committee met once or twice per year to monitor the conduct and progress of the trial and to address any safety issues. London Wandsworth Research Ethics Committee granted national research ethical approval (reference number 08/H0803/177), with additional local approvals for each recruitment centre.

Trial management

The trial was co-ordinated from a central trial office located within St George’s, University of London (SGUL), with the day-to-day running supervised and organised by the trial manager and administrator. The trial was sponsored by SGUL and conducted in accordance with Good Clinical Practice guidelines. 89 The chief investigator/trial manager and administrator received Good Clinical Practice training. Monthly research staff meetings were held at SGUL. There were no stopping rules or plans for interim analysis.

The NCTU provided a web-based database and randomisation system and data management reports. The system was held on a secure server in the NCTU, had a full electronic audit trail and full back-ups of the database were made every 24 hours. All of the outcome data were entered directly into the online forms by the participants or researchers. The database included validation checks whereby responses not meeting expected criteria would be flagged so that data entry errors were minimised.

The National Institute for Health Research (NIHR) Primary Care Research Network adopted the study. For public and patient involvement in the study see Appendix 8.

The protocol49 has been published and includes a description of approved amendments made to the original protocol after the start of recruitment; details of these amendments are given in the following section.

Recruitment of participants

Recruitment took place between April 2009 and November 2012. Follow-up at 6 months after birth continued until January 2014. Figure 1 shows accrual for the original recruitment target of 866 participants, the revised target of 774 participants [agreed with the NIHR Health Technology Assessment (HTA) programme following approval of the trial extension] and the 789 participants who were actually randomised. Four women were excluded post randomisation. Two women (PA group) were enrolled twice in sequential pregnancies and their second enrolment was removed. The other two women (control group) were excluded because they were found to be ineligible at their baseline visit before any data were collected and had been randomised erroneously. When ineligible participants are mistakenly randomised into a trial it is acceptable, within an ITT approach, to exclude their data without risking bias. 91 Table 4 shows the recruitment numbers for each centre for the final sample size of 785 participants included in the analysis. As shown in Table 5, over two-thirds of participants were recruited through midwife referral and over one-quarter through direct calling, through extracting contact information from the PAS. Fewer than 4% of participants were recruited using the other methods combined.

FIGURE 1.

Cumulative trial recruitment.

The CONSORT diagram (Figure 2) shows the flow of participants through the study to the primary end point at the end of pregnancy. Around 8100 women were recorded as smokers in the PAS. Of these, approximately one-quarter were uncontactable, just over one-quarter said that they were not interested in participating and just under one-third did not meet the inclusion criteria. A breakdown of the reasons for excluding participants is presented in Table 6. The main reason for exclusion was ‘reports smoking less than one cigarette a day’. Overall, of the 8096 women recorded as smokers at their first antenatal visit, 9.7% (785) were included in the ITT analysis. Of 785 pregnancies, 774 were singleton pregnancies, 10 were twin pregnancies and one was unknown as the woman withdrew consent.

FIGURE 2.

Numbers of participants who were enrolled in the study and included in the primary analysis. The participants lost to follow-up included some who had fetal or infant loss and were not assessed for smoking status.

Baseline characteristics

Participants in the two groups had similar baseline characteristics (Table 7). Women were recruited to the trial at a mean gestational age of 16 weeks and on average they were 27 years of age. Over half (53.6%) were smoking at least 10 cigarettes a day. By self-report, 70% were achieving the recommendation of 150 minutes a week of MVPA. 27,28

Follow-up rates

At 4 weeks after the quit day, 316 (80.6%) women were successfully followed up in the PA group and 319 (81.2%) in the control group. At the end of pregnancy follow-up, 587 (74.8%) women were assessed before the birth and 198 (25.2%) were assessed after the birth. The overall follow-up rate for the primary outcome at the end of pregnancy (see Figure 2) was 88.8% (697 participants) and there was no evidence of a significant difference in the follow-up rate between study groups. Of the 88 participants (11.2%) who did not complete the assessments necessary for analysis of the primary outcome, 43 (48.9%) were known to have smoked from the follow-up assessments (PA group n = 19, control n = 24). In addition, of the remaining 45 participants (5.7% of total), 24 had a fetal or infant death and were not asked about smoking status, and one individual withdrew consent; all were assumed to be smoking at the end of pregnancy.

Rates of biochemical validation of smoking status

For the majority of participants who reported that they were not smoking, biochemical validation was obtained. At the end of pregnancy, validation rates for smoking status were 71.4% (30/42) in the PA group and 56.8% (25/44) in the control group (p = 0.158); at 4 weeks, the rates were 98.0% (50/51) and 100.0% (61/61) respectively (p = 0.272).

Attendance at treatment sessions and compliance with the physical activity intervention

Participants attended a median of four of 14 treatment sessions in the PA group and three of six in the control group (Table 8). For the PA group compared with the control group, there was a 33% (95% CI 14% to 56%), 28% (95% CI 7% to 52%) and 36% (95% CI 12% to 65%) significantly greater increase in self-reported minutes of MVPA from baseline to 1 week, 4 weeks and 6 weeks respectively (see Table 8 and Figure 3). There was a decrease in self-reported minutes of PA at the end of pregnancy and at 6 months relative to baseline for both groups.

FIGURE 3.

Comparison^a of predicted self-reported levels of MVPA on a log-scale in the trial groups at different time points (bars represent CIs). EOP, end of pregnancy, FU, follow-up; PQ, post quit. a, Prediction on log scale from the mixed-effect model.

Of 90 participants asked to wear an accelerometer, 78 (86.7%) provided valid data (n = 37 PA group), 10 provided insufficient data and two were technical failures. Participants providing accelerometer data had similar baseline characteristics to those in the total sample. The majority (72%) had valid accelerometer data for at least 4 days. During the week of accelerometer wear, the median number of minutes of MVPA per day by self-report and according to accelerometer data was 38.2 [interquartile range (IQR) 25.4–54.5] and 7.8 (IQR 0–16.5) respectively. In total, 87% of participants self-reported higher levels of MVPA compared with the accelerometer data. Self-reports of minutes of MVPA per day were not significantly correlated with the accelerometer data (Spearman’s rho = 0.133, p = 0.247). Consistent with the correlation, using a Bland–Altman plot the mean difference between the self-report data and the accelerometer data for MVPA was 26.85 (95% CI 20.81 to 32.88) minutes (Figure 4). The median number of minutes of MVPA according to accelerometer data, when including only bouts of at least 10 minutes, was very similar for the PA group [7.5 (IQR 0–15.5)] and the control group [8.0 (IQR 0–16.2)] (p = 0.816). When including all MVPA, irrespective of duration, median activity levels per day tended to be higher for the PA group [38.00 (IQR 20.00–52.60)] than the control group [31.17 (IQR 19.00–44.10)], although this difference did not reach significance (p = 0.538).

FIGURE 4.

Bland–Altman plot of the difference between self-reported MVPA and accelerometer data for MVPA against the average of these measures (n = 78) showing the distribution of differences and how they relate to the average. Limits of agreement (reference range for difference) –26.698 to 80.390; range 0.714–72.286.

Only 28 women reported receiving face-to-face behavioural support for smoking cessation besides that offered in the trial and 60 women reported using NRT; the numbers reporting this support were similar in the two groups.

Smoking abstinence and reduction rates

There was no significant difference in smoking abstinence rates or smoking reduction rates between the two groups at follow-up (Table 9). The rate of validated continuous abstinence at the end of pregnancy (primary outcome) was 7.7% in the PA group and 6.4% in the control group (OR for PA group, adjusted for centre only, 1.21, 95% CI 0.70 to 2.10). At 4 weeks the validated abstinence rate was 12.8% in the PA group and 15.5% in the control group (OR, adjusted for centre only, 0.79, 95% CI 0.53 to 1.18). At 6 months postnatally the self-reported abstinence rate was 6.1% in the PA group and 4.1% in the control group (OR, adjusted for centre only, 1.55, 95% CI 0.81 to 2.97). Fully adjusted analyses yielded similar findings (at end of pregnancy: OR 1.37, 95% CI 0.78 to 2.41). The sensitivity analyses showed that the observed effect size and its statistical significance were independent of the influence of missing data for the primary outcome. There was no significant interaction between baseline self-reports of MVPA (< 150 vs. ≥ 150 minutes per week) and the treatment effect for the primary outcome [logistic regression model adjusted for site only: likelihood ratio test (LR chi-squared) = 2.31, p = 0.129; similar results were found for the fully adjusted model].

Withdrawal symptoms and urge to smoke

Tables 10 and 11 show the withdrawal symptom and urge to smoke scores, respectively, at baseline and 1 week post quit. When controlling for baseline score and treatment centre there was no significant group difference in total withdrawal score or urge to smoke score at 1 week post quit. When withdrawal symptoms were examined individually there were still no significant group differences.

Confidence for participating in physical activity and stopping smoking

Tables 12 and 13 present the scores for confidence for participating in PA and for stopping smoking, respectively, at baseline and 1 week post quit. When controlling for baseline score and treatment centre, ratings for confidence for participating in PA were significantly higher at 1 week post quit in the PA group than in the control group. When making the same adjustments, there was no significant group difference in confidence for quitting smoking at 1 week post quit.

Birth outcomes

Table 14 shows outcomes for singleton births. These outcomes were very similar between the two study groups except that there were significantly fewer deliveries by caesarean section in the PA group than in the control group (21.3% vs. 28.7%). Analyses that included twin births gave very similar findings.

Adverse events

There were similar numbers of AEs and SAEs in the two groups (Table 15). The total number of women or their infants who had at least one AE or SAE was 217 (55.4%) in the PA group and 219 (55.7%) in the control group (OR 0.99, 95% CI 0.75 to 1.32). The full breakdown of the data for less frequent AEs is provided in Appendix 9.

Maternal depression

All 784 participants provided EPDS data at baseline, with 383 (48.9%) and 279 (35.6%) participants providing these data at the end of pregnancy and at 6 months postnatally, respectively. The baseline characteristics of the subsamples used for the EPDS analysis at the two follow-up points were similar to those for the total trial sample. The baseline characteristics of the two trial groups were also similar in the subsamples at the end of pregnancy and at 6 months postnatally.

In both models the EPDS score was significantly higher in the PA group than in the control group at the end of pregnancy (Table 16). At this time there was a mean increase in EPDS score of 0.4 in the PA group and a mean reduction in EPDS score of 0.5 in the control group (mean difference between groups in fully adjusted model 0.95, 95% CI 0.08 to 1.83). When examining the data separately for end of pregnancy outcomes before and after the birth the findings were very similar. There was no significant difference in EPDS score between the groups at 6 months’ follow-up (fully adjusted mean difference 0.37, 95% CI –0.59 to 1.33).

Maternal weight

The combined sample for the analysis of maternal weight at the end of pregnancy was 271 participants, with 140 (51.7%) providing maternal weight before the birth and 131 (48.3%) providing maternal weight after the birth. The characteristics of these two subsamples (i.e. maternal weight provided before and after birth) were comparable with those of the main trial sample except that the rate of continuous smoking abstinence at the end of pregnancy was higher in both the subsample with maternal weight provided before delivery [n = 35 (25.0%)] and the subsample with maternal weight provided after delivery [n = 14 (10.7%)] than in the main sample [n = 55 (7.0%)]. In the combined sample at the end of pregnancy the characteristics were very similar between the two randomisation groups except that there was an approximately 1-kg difference in weight at early pregnancy between the two groups and therefore the results were adjusted for weight at early pregnancy. Rates of smoking abstinence were higher in the PA group than in the control group at the end of pregnancy [n = 29 (21.2%) vs. n = 20 (14.9%)] but this difference was not significant. When comparing the characteristics between the randomisation groups separately for the subsamples providing maternal weight before and after birth the findings were very similar.

At the end of pregnancy the mean GWG (subsample providing maternal weight before birth) in the PA group was 12.4 kg [standard deviation (SD) 6.1 kg] and in the control group was 11.2 kg (SD 7.0 kg) and the postnatal weight retention (subsample providing maternal weight after birth) was 4.7 kg (SD 7.2 kg) and 4.9 kg (SD 7.3 kg), respectively. With the basic adjustment the mean difference in GWG and postnatal weight retention for the PA group compared with the control group was 1.08 kg (95% CI –1.08 kg to 3.23 kg) and 0.11 kg (95% CI –2.27 kg to 2.49 kg), respectively (Table 17). In the fully adjusted model these mean differences were reduced and neither of them achieved statistical significance. The estimated mean weight change per gestation week was 0.31 kg (SD 0.15 kg) in the PA group and 0.27 kg (SD 0.17 kg) in the control group, and the difference was not significant. In contrast, the difference in estimated mean weight change per gestation week between the groups was significantly lower for those classed as obese at baseline than for those classed as non-obese at baseline (–0.12 kg, 95% CI –0.21 kg to –0.03 kg; p = 0.010). There was no significant interaction between the effect of PA compared with the control and whether or not the individual was obese at baseline (p = 0.599), such that, in those who were not obese, the difference between the PA group and the control group was 0.03 kg (95% CI –0.03 kg to 0.08 kg) and in those who were obese the difference between the PA group and the control group was 0.06 kg (95% CI –0.09 kg to 0.21 kg).

According to Institute of Medicine guidelines88 the overall proportion of women with excessive GWG was 38.6%. The risk of excessive GWG was slightly higher in the PA group than in the control group (fully adjusted OR 1.35, 95% CI 0.63 to 2.88), but the difference was not significant (see Table 17). The risk was significantly higher for the overweight and obese categories than for the healthy weight category (overweight: fully adjusted OR 4.0, 95% CI 1.69 to 9.58; obese: fully adjusted OR 3.29, 95% CI 1.20 to 8.99). Those in the underweight category were less likely to gain excessive gestational weight than those with a healthy weight but this was not significant (fully adjusted OR 0.57, 95% CI 0.06 to 5.31). In sensitivity analyses, when adjusted for the effect of baby’s weight and rates of smoking abstinence, the mean difference in the weight change between the two groups was slightly changed and the effect of baby’s weight was significant in the regression models (see Table 17).

Introduction

Several studies have investigated the potential cost saving of smoking cessation interventions in pregnancy92 but only one study could be identified that has used empirical data on costs to calculate the incremental cost-effectiveness of these interventions. 93 This chapter reports on an economic evaluation conducted alongside the LEAP trial to address the cost-effectiveness of a PA intervention plus behavioural support compared with behavioural support alone.

The objectives were:

1. to compare the costs associated with the control and intervention strategies

2. to estimate the consequences of these alternatives

3. to assess the cost-effectiveness of the PA intervention used in addition to behavioural support for smoking cessation at the end of pregnancy.

Methods

Results

In the PA group 30 of 391 women (7.7%) were abstinent from smoking with biochemical validation from the quit date to the end of pregnancy; the equivalent figure for the control group was 25 of 393 women (6.4%), a non-significant difference (see Chapter 3, Results).

Table 19 reports the health-care resources utilised in the two arms of the trial. Quantities of services were multiplied by the relevant unit costs in Table 18 to calculate the resource use for each woman. Maternal antenatal and postnatal hospital admissions and admissions to neonatal care were similar in the two groups. With regard to mode of birth, the PA group underwent more assisted and spontaneous vaginal births and the control group underwent more caesarean sections and miscarriages. A chi-squared test including all four modes of birth showed a significant difference between the groups (χ² = 8.7, p < 0.035). On average, the PA group attended 5.3 of 14 treatment sessions offered and the control group attended 3.5 of six sessions offered. The total duration of treatment sessions was around twice that for the PA group compared with the control group.

The point estimate results shown in Table 20 suggest that the PA intervention is somewhat less costly than the control intervention, by a margin of £35 per participant, but produces a small increase in the expected number of quitters. Taken in isolation, this suggests that the PA intervention should be adopted, regardless of the decision-maker’s willingness to pay for an additional quitter. However, this result must be interpreted with great caution because there is considerable statistical uncertainty, as reflected in the scatterplot (Figure 5), which shows that neither the cost difference nor the effect difference is statistically significant as there are an appreciable number of points in each quadrant. This is confirmed by the cost-effectiveness acceptability curve (Figure 6), which shows that the proportion of bootstrapping replications deemed cost-effective does not exceed 80% for any willingness-to-pay threshold up to £50,000 per additional quitter.

FIGURE 5.

Cost-effectiveness scatterplot.

FIGURE 6.

Cost-effectiveness acceptability curve.

The ‘stripey’ effect in the cost-effectiveness scatterplot (see Figure 5) reflects the discrete nature of the outcome measure. One additional quitter selected in the bootstrapping moves the point to the next stripe. It appears that the estimated cost saving from the PA intervention is driven by the increased use of health-care resources in the control group, particularly with regard to the rate of caesarean section.

The results were calculated assuming that the costs for treadmill and midwife training would be based on a typical hospital size. The effect of these costs is negligible compared with other costs in the analysis and so varying this assumption makes no appreciable difference to the results. Table 21 presents illustrative results showing the effect on overall costs and the difference in costs of two possible scenarios. In the first scenario the per-participant fixed costs were halved, assuming that the treadmill and training would apply to twice as many participants (116 per year instead of 58). In the second scenario the per-participant fixed costs were doubled, assuming that the treadmill and training would apply only to half as many participants (29 per year).

In a subgroup analysis, when repeating the calculations for the two categories of age and FTCD score, the point estimates no longer show PA dominating the control, but the bootstrapped results still occupy all four quadrants of the cost-effectiveness plane and, thus, it cannot be concluded that the intervention is cost-effective (or not) for any of these subgroups. Further details of the subgroup analysis are provided in Appendix 10.

Summary

Costs associated with the control and PA interventions were compared and the consequences of these alternatives were estimated. The main results of the cost-effectiveness analysis can be summarised as follows:

• Mean costs were £35 per participant lower in the PA group than in the control group. This results from the fact that the mean intervention costs were £27 higher in the PA group than in the control group but health-care resource use costs were £62 higher in the control group. An important contributory factor to this result was the difference in rates of caesarean section (25% control group vs. 19.7% PA group).

• Taken in isolation, the reduction in mean costs and increase in quit rate suggest that the PA intervention should be adopted, regardless of the decision-maker’s willingness to pay for an additional quitter. However, this result must be interpreted with great caution because there is considerable statistical uncertainty, as reflected in the scatterplot and accompanying cost-effectiveness acceptability curve.

• Subgroup analysis was performed by age and dependency score. In each subgroup analysis there was still considerable uncertainty in both the incremental costs and the difference in quit rate.

Smoking outcomes

The trial showed that, among women who were recruited at 10–24 weeks of pregnancy, supplementing behavioural support with a PA intervention was no more effective than behavioural support alone in promoting smoking cessation at the end of pregnancy. At this time, the self-reported continuous smoking abstinence rate, with biochemical validation, was 1.3% higher among those in the PA group than among those in the control group, but this difference was not statistically significant.

Strengths and limitations

This is the first trial to assess the effect of a PA intervention on smoking cessation during pregnancy. Limitations of previous trials of PA interventions for smoking cessation were overcome22 through offering an intensive PA intervention (with support to increase PA as an aid for smoking cessation in addition to supervised exercise sessions), assessing and validating PA in both groups, using a robust outcome of continuous smoking abstinence and offering a pragmatic intervention that could be readily integrated into routine health care in the NHS.

This study was approximately twice as large as previous RCTs investigating PA interventions for smoking cessation delivered via face-to-face support. 22 As anticipated, around 10% of pregnant women recorded as smokers at their first antenatal booking visit were recruited. However, the overall number of women recorded as smokers was lower than anticipated; therefore, despite extending the recruitment period, we recruited only 91% of the target population of 866 (i.e. n = 789). Future studies may wish to aid recruitment by focusing on regions of the UK that have particularly high rates of smoking in pregnancy. Quit rates were lower than anticipated in our power calculation, which will also have reduced the power of the study. The CI around the OR for the effect of the intervention shows the level of precision of this estimate, given the achieved sample size, and implies that we cannot rule out up to a twofold increase in abstinence at the end of pregnancy in the intervention group compared with the control group at the upper end of the CI, although our point estimate suggests an effect of the intervention that is unlikely to be clinically meaningful. Additionally, at the 4-week follow-up point quit rates were higher and CIs narrower, and the difference in abstinence between the treatment group and the control group was minimal, even at the upper end of the CI. Thus, despite not reaching the target sample size, our results suggest that there is unlikely to be a clinically relevant difference between the groups. Quit rates were lower than in previous pregnancy trials with less rigorous outcome measures but were similar to those in studies using comparable outcome measures. 12 For example, two previous UK-based large trials of a smoking cessation intervention during pregnancy observed quit rates at the end of pregnancy of 7.6% and 7.8% in a group receiving only behavioural support for smoking;95,102 similarly, we observed a quit rate of 6.4% for this group.

Our finding of a lack of an effect of a PA intervention on smoking abstinence is consistent with most previous trials of PA for smoking cessation among non-pregnant smokers. However, these trials had low statistical power or used insufficiently intense interventions. 22 One adequately powered trial showed that regular, supervised, vigorous-intensity PA was effective for smoking cessation in women. 23 However, such an intensive intervention is not likely to be clinically suitable or appealing to most pregnant smokers and our finding remains that a pragmatic intervention based on moderate-intensity exercise was not effective for aiding smoking cessation during pregnancy.

Secondary outcomes

Economic evaluation

The total mean cost was £35 per participant lower in the PA group than in the control group. This results from the mean health-care resource use costs being £62 higher in the control group than in the PA group and the mean intervention costs being £27 higher in the PA group than in the control group. Control group health-care costs were higher mostly because of the higher rate of caesarean sections in this group (25% control group vs. 19.7% PA group). Considering the reduction in mean cost and increase in quit rate for the PA group compared with the control group, it could be recommended that the PA intervention be used as an aid for smoking cessation regardless of the decision-maker’s willingness to pay for an additional quitter. However, these results must be interpreted with extreme caution as there is considerably statistical uncertainty, as reflected in the scatterplot and accompanying cost-effectiveness acceptability curve.

Previous studies have investigated the cost-effectiveness of smoking cessation interventions during pregnancy. 92 However, these mainly US-based studies are limited by not including data for infant outcomes and therefore cannot easily be compared with this study. One previous study (the SNAP trial96), based in the UK, has considered infant outcomes and conducted a similar analysis as used in the present study. The SNAP trial assessed the cost-effectiveness of NRT for smoking cessation. The authors reported that the total mean cost of the intervention was £90.81 higher in the NRT group than in the usual-care group. While bearing in mind the high levels of statistical uncertainty present in both studies, this suggests that the PA intervention may be at least as cost-effective, if not more so, than the NRT intervention.

The National Institute for Health and Care Excellence recommends that health outcomes should be measured in quality-adjusted life-years (QALYs) to facilitate comparisons between different health-care programmes. 94 Ideally, the value to society of each successful quitter in the LEAP trial would be estimated in QALYs but no method has been found for doing this that can be considered robust and reliable for use with pregnant smokers. QALYs are commonly calculated from generic health-related quality of life tools (e.g. the European Quality of Life-5 Dimensions or EQ-5D), which may be inappropriate for use in pregnant populations. Quality-of-life studies using generic measures have demonstrated that changes in quality of life, particularly declining physical functioning and vitality, occur over the course of pregnancy. 109 These substantial changes in quality of life may mask any potential short-term quality-of-life gains from smoking cessation. Moreover, existing models ignore QALY benefits to the fetus and do not take into account maternal morbidity during pregnancy. 92,93,110 It was anticipated that a model and systematic review being developed for another study at the University of Nottingham would be suitable for this purpose but this work has yet to be completed. However, once an appropriate economic model that values smoking in pregnancy is available we will use this to estimate cost-effectiveness in QALYs.

Interpretation and generalisability of the results for smoking outcomes

The trial had reasonably broad inclusion criteria and few exclusion criteria, and, therefore, the results are likely to be generalisable to most pregnant smokers, although the women were highly physically active in both groups at baseline. At baseline, compared with a survey of pregnant smokers similarly recruited in antenatal hospitals,111 participants were approximately twice as likely to report achieving the recommended 150 minutes per week of MVPA; this may be because more active women were attracted to a trial offering a PA intervention. Low attendance may have affected efficacy, with the PA group attending a median of only four (of 14) sessions. Low attendance was not explained by the two potentially treatment-related AEs in the PA group and there was no indication that the PA intervention increased the overall incidence of AEs. The vast majority of women who failed in their quit attempt stopped attending, suggesting that it was failure to quit that led to low attendance rather than low attendance leading to failure of the attempt. There was no evidence for the intervention influencing processes that might aid cessation, such as confidence for quitting, urges to smoke or withdrawal symptoms.

Bias in outcome ascertainment is unlikely to explain the findings as follow-up rates were equally high in the groups and the effect size was independent of the influence of missing outcome data. Fewer than 10% of participants reported using non-study behavioural support or NRT, with similar usage in the groups. Therefore, it is doubtful that this influenced the results.

Although the intervention group reported significantly higher PA levels throughout pregnancy relative to the control group, the self-reported PA scores in the control group were also relatively high at follow-up. For example, at the end of pregnancy the intervention group reported completing a median of 155 minutes of MVPA per week (22 minutes per day) and the control group reported completing a median of 140 minutes of MVPA per week (20 minutes per day). This suggests that some intervention contamination might have occurred in the control group and consequently there was an insufficient difference in PA levels between the groups to show an appreciable effect on depression outcomes. It is not clear why the control group participated in more PA than might be expected. One explanation could be that an atypical motivated sample of participants was recruited who were keen to be active regardless of their random group allocation; the high baseline PA scores would suggest this explanation is a possibility as participants were already achieving around 30 minutes per day of MVPA at baseline. The only previous trial of exercise for smoking cessation showing a long-term benefit for abstinence excluded more active women,23 and in less active women the LEAP intervention might have had more positive effects. However, there was no evidence to suggest that the treatment effect for the primary outcome was influenced by baseline levels of PA. Moreover, as participants were recruited from routine health-care settings, it seems likely that if the exercise intervention was offered as part of routine care women would be recruited who are active at a similar level to those taking part in the current trial.

It is also important to consider that, as previously observed,112–114 our accelerometer correlational analysis and Bland–Altman plot suggest that participants overestimated their self-reported PA levels. Moreover, participants could not be blinded to treatment allocation and the higher self-reported rates of activity in the PA group may be the result of a social desirability bias, such that there is a tendency to over-report PA levels with the belief that this will be positively viewed by the researchers. In addition, in the subsample using the accelerometer, accelerometer-derived PA levels were similar in the two trial groups, although when including all MVPA (rather than restricting it to bouts of > 10 minutes) the PA group tended to report more PA per week than the control group, which suggests that most increases in activity were likely to be in sporadic bouts lasting < 10 minutes. The accelerometer data were collected for only 10% of participants and therefore this finding must be treated with caution; however, it is possible that the non-significant effects for smoking abstinence emerged partly because the PA group failed to adhere fully to the behavioural goals of the intervention. If this is the case, it is unlikely that such an intervention offered in routine health care would raise PA levels sufficiently to have an impact on smoking cessation. On the other hand, although attendance at treatment sessions was low, the PA counselling may have increased PA even without attendance at all scheduled exercise sessions. A less pragmatic and more efficacy-oriented trial (e.g. with greater incentives to complete the exercise, such as financial incentives) with accelerometer measurement in the whole sample would be needed to establish whether or not PA per se aids smoking cessation during pregnancy.

Conclusion

During pregnancy, offering an intervention combining supervised exercise and PA counselling did not add to the effectiveness of behavioural support for smoking cessation. There was no evidence that the PA intervention increased AEs or had a harmful effect on birth outcomes, and there was some evidence that the PA intervention resulted in fewer caesarean sections. There was no evidence for the intervention reducing maternal depression or weight gain.

Recommendations for research (in priority order)

1. It is not recommended to fund further large-scale trials of PA for smoking cessation until much less expensive observational studies have been conducted to provide promising leads, for example related to the populations most suitable for such interventions and methods for increasing PA adherence.

2. Reasons for pregnant smokers’ low levels of attendance at supervised PA sessions should be investigated; findings could be used to increase attendance rates. For example, following on from recent work on barriers to PA,115–117 further research is needed to explore barriers to attendance and to PA adherence during pregnancy, and to assess whether or not these barriers vary during different stages of pregnancy and among women with different comorbidities, including gestational diabetes and obesity.

3. Further methods of increasing PA adherence among pregnant smokers need to be developed and tested. For example, financial incentives have shown some benefit for aiding smoking cessation in this population and they may be used in combination with PA to increase both attendance at exercise sessions and smoking cessation. 12 In addition, interventions are needed that provide regular prompts to remind women to exercise (e.g. text messages or brief telephone calls); such interventions have been successfully piloted with young women but not yet with pregnant women. 118

4. The reasons why few inactive pregnant smokers were attracted to a PA trial need to be identified and methods are needed to attract these less active pregnant smokers.

5. Studies are needed to establish whether or not the previously reported finding of a short bout of PA reducing cigarette cravings in pregnant smokers is a robust finding. So far, only one study has investigated this issue. If it is a robust finding, interventions need to be developed that can translate this benefit into prevention of smoking relapse.

6. There was no evidence of beneficial effects on maternal weight gain or depression. Studies are needed that focus on women who are at risk of higher maternal weight gain and on women who have high levels of depression at baseline.

7. Among pregnant smokers there was no evidence for a PA intervention having an added benefit for smoking cessation beyond that of usual care. However, it is possible that in some circumstances a PA programme alone may be more practical and may aid smoking cessation, and this needs to be assessed.

8. There were significantly fewer deliveries by caesarean section in the PA group than in the control group. Further studies are needed to replicate this finding and to explore the underlying mechanisms.

Implications for health care

There was no evidence that offering regular, supervised exercise and PA consultations, in addition to routine smoking cessation support, to women following their first antenatal visit is effective for aiding smoking cessation. Therefore, PA is not currently recommended for smoking cessation during pregnancy. One study showed that PA acutely reduces cigarette cravings in abstinent pregnant smokers29 and, among smokers in general, PA is recommended for reducing these cravings. 21,22 In the present study there was no evidence for the PA intervention moderating cravings/urges to smoke, but it is possible that there are some acute benefits of PA on reducing cravings during pregnancy and the recommendation to use PA to manage cravings acutely remains for all smokers, including those who are pregnant. The PA intervention did not show any benefit for reducing maternal depression; in fact, there was a slight increase in depression in the PA group and therefore the intervention cannot be recommended for antenatal or postnatal depression in women attempting to quit smoking during pregnancy. There was no evidence for an effect of the intervention on maternal weight gain and therefore the intervention cannot be currently recommended for moderating this weight gain. There was no evidence of an increased level of AEs in the PA group and there was some evidence for a reduced incidence of caesarean sections in the PA group; therefore, in line with current guidance, PA remains indicated for general health benefits in pregnancy, including among pregnant smokers.

Trial team

Chief investigator: Michael Ussher.

Coapplicants/Trial Management Group: Paul Aveyard, Isaac Manyonda, Sarah Lewis, Robert West, Beth Lewis, Bess Marcus, Adrian H Taylor, Pelham Barton, Tim Coleman.

Trial managers: Michael Ussher, Noura Hamdi (year 1 only).

Trial administrator: Mary Apps.

Trial Steering Committee: Jim Thornton (Chair), Ann McNeil, Tim Coleman, Michael Ussher, Sue Cooper, Kim Watts, Serena Cox (patient and public involvement/lay member).

Statisticians: Sarah Lewis, Muhammad Riaz.

Data cleansing and database preparation: Sarah Kerry.

Health economists: Pelham Barton, Holly Essex.

Researchers: Julie Fuller, Maggie Hart, Bettina Wanninkhof, Ilia Papachristou, Gail Harding, Sarah Cleary, Ory Bolooki, Rachel Lex, Beth Steff, Zoe Magrath, Tracey Kilbane, Janet Brown, Caroline Dixon, Noura Hamdi.

Principal investigators (in recruiting centres): Isaac Manyonda (St George’s Healthcare NHS Trust), Mark Johnson (Chelsea and Westminster Hospital NHS Foundation Trust), Andrew Shennan (Guy’s and St Thomas’ NHS Foundation Trust), Raj Rai (Imperial College Healthcare NHS Trust), Ranee Thakar (Croydon Health Services NHS Trust), Hassan Shehata (Epsom and St Helier University Hospitals NHS Trust), Nick Anim-Nyame (Kingston Hospital NHS Foundation Trust), Katie Yiannouzis (King’s College Hospital NHS Foundation Trust), Maureen Royds-Jones (Surrey and Sussex Healthcare NHS Trust), Gill Perks (Medway Foundation Trust), Joanna Girling (West Middlesex University Hospital NHS Trust) and Simon Cunningham (Mid Cheshire Hospitals NHS Foundation Trust).

NCTU management: Dan Simpkins.

Cotinine analysis: Salimetrics Europe Ltd (Newmarket, UK), Dr Agnes Ernst.

We would also like to thank the following for their contributions to the study: Sherma Garcia and Tracey MacCormack for their help with data collection, and Dr Judith Ibison and Professor Andy Kent for their advice about the measurement of depression and interpretation of the findings for depression.

Contributions of authors

All authors made substantial contributions to the conception and design of the study and/or the acquisition of data and/or the analysis and interpretation of data, as listed below. All authors were involved in drafting the manuscript or revising it critically for important intellectual content. All authors approved the final version.

Michael Ussher (LEAP trial chief investigator/trial manager, Professor of Behavioural Medicine) was involved in the design, conduct, data acquisition, analysis and report-writing phases of the trial.

Sarah Lewis (Professor of Medical Statistics) was involved in the design, conduct, analysis and report-writing phases of the trial.

Paul Aveyard (Professor of Behavioural Medicine) was involved in the design, conduct, analysis and report-writing phases of the trial.

Isaac Manyonda (Professor and Consultant in Obstetrics and Gynaecology) was involved in the design, conduct and report-writing phases of the trial.

Robert West (Professor of Health Psychology) was involved in the design, conduct and report-writing phases of the trial.

Beth Lewis (Associate Professor, Behavioural Aspects of Physical Activity) was involved in the design, conduct and report-writing phases of the trial.

Bess Marcus (Professor of Psychiatry and Human Behaviour and Community Health) was involved in the design, conduct and report-writing phases of the trial.

Muhammad Riaz (Research Fellow in Medical Statistics) was involved in the data cleansing, analysis design, analysis and report-writing phases of the trial.

Adrian H Taylor (Professor of Exercise and Health Psychology) was involved in the design, conduct and report-writing phases of the trial.

Pelham Barton (Reader in Mathematical Modelling) was involved in the design, conduct, analysis and report-writing phases of the trial specifically for the economic analysis.

Amanda Daley (Senior Lecturer in Behavioural Medicine) was involved in the design, analysis and report-writing phases of the trial specifically for the outcomes related to depression and maternal weight.

Holly Essex (Research Fellow in Health Economics) was involved in the design, analysis and report-writing phases of the trial specifically for the economic analysis.

Dale Esliger (Senior Lecturer in the Measurement of Physical Activity) was involved in the design, conduct, analysis and report-writing phases of the trial specifically for accelerometer data outcomes.

Tim Coleman (Professor of Primary Care) was involved in the design, conduct, analysis and report-writing phases of the trial.

Publications

Ussher M, Aveyard P, Manyonda I, Lewis S, West R, Lewis B, et al. Physical activity as an aid to smoking cessation during pregnancy (LEAP) trial: study protocol for a randomized controlled trial. Trials 2012;13:186.

Ussher M, Lewis S, Aveyard P, Manyonda M, West R, Lewis B, et al. Physical activity for smoking cessation in pregnancy: randomised controlled trial. BMJ 2015;350:h2145.

Data sharing statement

The guarantor (MU) is willing to examine all requests for the full data set after a period of 3 years from the date of this publication. Participants did not give consent for data sharing but the data are anonymised and the risk of identification is low.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.

Instructions for carrying out the 7-day physical activity recall interview

1. Label the days on the recall table (day 1 is yesterday).

2. Introduce interview: Explain that you will be asking about their physical activity, starting with yesterday and working backwards through the previous 7 days.

3. Define segments of the day: Say that you will be asking separately about activities for the morning, afternoon and evening.

4. Describe the type of activity that you are interested in: Explain that you are interested in any work, household or leisure activities lasting at least 10 minutes that are at least at a level of intensity that makes them breathe slightly harder than normal, makes them warmer and makes them aware that their heart is beating faster. Say that you are not interested in light activities such as desk work, strolling or light housework.

5. Ask about activities: For each segment of the day, ask the woman to recall activity episodes lasting at least 10 minutes and record the information in the table (round to the nearest 5 minutes). For each activity record the type of activity (using type codes) and duration of activity in minutes (e.g. ‘W 10’ denotes a 10-minute walk). For each day begin by asking ‘What did you do and where did you go on that morning’.

6. Record the number of minutes of activity for each day and the total minutes for the week. Enter the latter figure into the database.

Baseline questionnaire

(PDF download)

Additional questions asked post quit day only

(PDF download)

Antenatal complications and birth out comes

(PDF download)

Serious adverse events