Financial incentives for smoking cessation in pregnancy
Randomised controlled trial.
“Financial incentives for smoking cessation in pregnancy: randomised controlled trial” by David Tappin and colleagues (BMJ 2015;350:h134)
Objective—To assess the efficacy of a financial incentive added to routine specialist pregnancy stop smoking services versus routine care to help pregnant smokers quit.
Design—Phase II therapeutic exploratory single centre, individually randomised controlled parallel group superiority trial.
Setting—One large health board area with a materially deprived, inner city population in the west of Scotland, United Kingdom.
Participants—612 self reported pregnant smokers in NHS Greater Glasgow and Clyde who were English speaking, at least 16 years of age, less than 24 weeks pregnant, and had an exhaled carbon monoxide breath test result of 7 ppm or more. 306 women were randomised to incentives and 306 to control.
Interventions—The control group received routine care, which was the offer of a face to face appointment to discuss smoking and cessation and, for those who attended and set a quit date, the offer of free nicotine replacement therapy for 10 weeks provided by pharmacy services, and four, weekly support phone calls. The intervention group received routine care plus the offer of up to £400 (€560; $600)of shopping vouchers: £50 for attending a face to face appointment and setting a quit date; then another £50 if at four weeks’ post-quit date exhaled carbon monoxide confirmed quitting; a further £100 was provided for continued validated abstinence of exhaled carbon monoxide after 12 weeks; a final £200 voucher was provided for validated abstinence of exhaled carbon monoxide at 34-38 weeks’ gestation.
Main outcome measure—The primary outcome was cotinine verified cessation at 34-38 weeks’ gestation through saliva (<14.2 ng/mL) or urine (<44.7 ng/mL). Secondary outcomes included birth weight, engagement, and self reported quit at four weeks.
Results—Recruitment was extended from 12 to 15 months to achieve the target sample size. Follow-up continued until September 2013. Of the 306 women randomised, three controls opted out soon after enrolment; these women did not want their data to be used, leaving 306 intervention and 303 control group participants in the intention to treat analysis. No harms of financial incentives were documented. Significantly more smokers in the incentives group than control group stopped smoking: 69 (22.5%) versus 26 (8.6%). The relative risk of not smoking at the end of pregnancy was 2.63 (95% confidence interval 1.73 to 4.01) P<0.001. The absolute risk difference was 14.0% (95% confidence interval 8.2% to 19.7%). The number needed to treat (where financial incentives need to be offered to achieve one extra quitter in late pregnancy) was 7.2 (95% confidence interval 5.1 to 12.2). The mean birth weight was 3140 g (SD 600 g) in the incentives group and 3120 (SD 590) g in the control group (P=0.67).
Conclusion—This phase II randomised controlled trial provides substantial evidence for the efficacy of incentives for smoking cessation in pregnancy; as this was only a single centre trial, incentives should now be tested in different types of pregnancy cessation services and in different parts of the United Kingdom.
Why do the study?
Smoking in pregnancy harms unborn babies and increases the risk of premature birth, miscarriage, stillbirth, and maternal complications. In the United Kingdom alone, an estimated 293 babies and infants die each year as a direct result of exposure to tobacco smoke during pregnancy. Encouraging women to quit is likely to save lives, improve the long term health of mothers and their children, and ultimately cut the cost of treating smoking related health problems, which can be life long. Quitting smoking is difficult, even in pregnancy, and the usual strategies such as nicotine replacement are not particularly effective. These authors are evaluating an alternative approach: to offer pregnant women financial incentives to stop smoking, in addition to routinely available smoking cessation services.
What did they do?
Tappin and colleagues designed a randomised controlled trial to test the effect of offering pregnant smokers up to £400 (€560; $600) worth of shopping vouchers to quit. The trial compared quit rates in 306 pregnant smokers who were offered the vouchers in addition to routinely available smoking cessation services and 306 pregnant smokers offered routinely available services alone, with no financial incentives.
The primary outcome was self reported abstinence from smoking at 34 to 38 weeks’ gestation. The authors corroborated women’s reports by testing for carbon monoxide in exhaled breath, and for cotinine in both urine and saliva. Cotinine is a metabolite of nicotine and a good biomarker for exposure to tobacco smoke.
Women were recruited in early pregnancy, and referred to national smoking cessation services which offered counselling, telephone support, and free nicotine replacement therapy for 10 weeks. Intervention women were also offered a series of incentives, starting with £50 of vouchers if they attended an appointment and set a quit date, further vouchers if they reported abstinence at 4 weeks and 12 weeks, then a final £200 if they were still not smoking at 34 to 38 weeks. At each stage, self report was corroborated using exhaled carbon monoxide, then later with cotinine concentrations in urine or saliva.
Although the primary aim of the trial was to compare women at the end of pregnancy, the authors also compared how many women in each group set a quit date, how many reported abstinence four weeks later, and how many quitters still reported abstinence six months after delivery. The authors recorded all birth weights, premature births, miscarriages, and still births from maternity records.
What did they find?
Women offered the shopping vouchers were significantly more likely than control women to be smoke free towards the end of pregnancy (69/306, 22.5% v 26/303, 8.6). The offer of vouchers more than doubled women’s chances of not smoking at 34 to 38 weeks (relative risk 2.63, 95% confidence interval 1.73 to 4.01), and the authors estimate that one extra woman would quit for every seven offered vouchers. The “number needed to treat” is a good indication of effectiveness. The lower the better—seven is relatively low for any intervention, including other drug or behavioural options for smokers trying to quit.
The offer of shopping vouchers also improved self reported quit rates at 4 weeks (133/306, 43% v 64/303, 21%, P<0.001), but had no effect on other prespecified secondary outcomes. Six months after delivery, a substantial proportion of quitters in both groups had relapsed (18/69, 26% of women in the incentives group and 11/26, 42% of controls). This analysis was post hoc, not part of the original trial plan. Cotinine tests on routine blood samples from 200 participants suggested that women with incomplete data—dropouts or women who failed to present for tests—were likely to be still smoking at the end of pregnancy.
The authors used the same blood samples to look for “gaming” whereby women might quit smoking for a few days before being tested for cotinine or carbon monoxide, then resume soon afterwards. From the limited data available, gaming looked rare. Eighty per cent of blood samples from women classified as quitting at 34-38 weeks confirmed that these women were truly abstinent.
What are the study’s strengths and weaknesses?
Randomised trials are near the top of the evidence hierarchy, and the only truly reliable way to test treatments, devices, and other interventions such as financial incentives.
Random allocation generates two (or more) very similar groups of people, so any differences between them at the end of a trial can be attributed to their allocated treatment. Here are some key strengths to look out for in RCTs, all present in Tappin and colleagues’ trial.
Size is important because it is closely related to power, and power is closely related to reliability. With 612 participants this trial is big enough and powerful enough (90% power) to rule in or rule out a clinically useful effect with confidence. The authors report a clear power calculation detailing the magnitude of effect they were looking for, and number of participants they needed for a reliable result. They recruited enough women and analysed enough outcomes to meet their prespecified requirements
Randomisation generates an allocation sequence which must be concealed. Otherwise, a researcher can see which treatment option will be allocated next, and can theoretically select a participant for that particular trial arm. Tappin and colleagues say they concealed allocation, but don’t say how.
Blinding or masking
Not to be confused with allocation concealment, blinding is important when measuring outcomes. In Tappin and colleagues’ trial, the researchers recording women’s smoking habits in late pregnancy (primary outcome) did not know which group the women were in, so had no opportunity to bias the results in favour of one group or another. This trial was single blind. In double blind trials, the participants are also unaware of their treatment group.
Intention to treat analysis
In this kind of analysis, all participants allocated to a particular treatment group are analysed together at the end of a trial, even if they drop out, fail to comply with treatment, switch groups, or otherwise deviate from the prespecified protocol. Tappin and colleagues faithfully analysed women within their allocated groups. They assumed that women who dropped out or failed to provide smoking data at the end of pregnancy were still smoking. Authors who don’t use intention to treat analysis should explain why.
Missing data are one of this trial’s main weaknesses. In all, 276 women out of 609 failed to report their smoking habits four weeks after their quit date. So the secondary results comparing quit rates at this stage may not be reliable.
The two post hoc analyses (comparing quit rates six months after delivery, and evaluating gaming using routine blood samples) also need confirming in future trials. Post hoc analyses are always exploratory and readers should treat them with caution. You can spot post hoc analyses by checking the trial register (cut and paste the registration number given at the end of the abstract into a search engine). If the outcome “quit rate 6 months after delivery” isn’t listed on the register, the authors must have decided to add it later. All randomised trials should be registered before the first patient is recruited. That way, readers can check whether the trial was carried out and reported faithfully, not manipulated in any way to bias findings or conclusions.
Finally, women’s responses to financial inducements may be culturally specific. It would be unwise to generalise these findings far beyond this particular area of Scotland, where smoking in pregnancy is concentrated among women on low incomes.
What do the findings mean?
Financial incentives such as shopping vouchers look like a promising way to encourage pregnant women to stop smoking. This trial suggests that the added incentive works better than Scotland’s routine cessation services alone and the authors plan a more definitive trial to test this approach further. Their findings also show just how hard it is to give up smoking. Less than one quarter of women managed to quit by the end of pregnancy, even when offered a substantial incentive worth up to £400. Controls fared even worse. Relapse rates were high following delivery, which is likely to reduce long term cost effectiveness. So, as Deirdre Murphy says in her linked editorial “incentives help, but not enough and not for long.”
Financial incentives are controversial. Society has yet to decide whether paying people to be more healthy is a fair use of public resources, or ethically suspect bribery and coercion of society’s poorest citizens.Alison Tonks, associate editor, The BMJ
Correspondence to: firstname.lastname@example.org
Competing interests: AT helps select research for publication in The BMJ, and was the handling editor for the research paper featured in this article and associated editorial by DJ Murphy.
Provenance and peer review: Commissioned; not externally peer reviewed.
- Tappin D, Bauld L, Purves D, et al. Financial incentives for smoking cessation in pregnancy: randomised controlled trial. BMJ 2015;350:h134.
- Murphy DJ. Financial rewards for pregnant smokers who quit. BMJ 2015;350:h297.
Cite this as: Student BMJ 2015;23:h1419