Katherine Everett-Murphy, Chronic Diseases of Lifestyle Unit, Medical Research Council, PO Box 19070, Tygerberg, 7505 Cape Town, South Africa. E-mail: email@example.com
Aim and objectives. To evaluate the effect of a smoking cessation intervention, based on best practice guidelines on the quit rates of disadvantaged, pregnant women in Cape Town, South Africa. Design. Quasi-experimental using a natural history cohort as a control group, consisting of women attending antenatal care in 2006 and an intervention cohort, attending the same clinics a year later. Setting. Four, public sector antenatal clinics in Cape Town staffed and managed by midwives. Population. Pregnant women of low socio-economic status. Methods. The natural history cohort received usual care, whilst the intervention cohort was offered self-help quit materials in the context of brief counseling by midwives and peer counselors. Smoking behavior was measured in early, mid and late pregnancy. The equivalence of the groups in terms of smoking profile, self-reported smoking and demographic variables was assessed at baseline. Main outcome measures. Quit rates measured by urinary cotinine towards the end of pregnancy (36–39 weeks gestation). Results. The two cohorts were comparable at baseline. The difference in quit rates between the two cohorts in late pregnancy was 5.3% (95% CI: 3.2–7.4%, p < 0.0001) in an intention to treat analysis. There was also a significant difference in reduction of smoking of 11.8% (95% CI: 5.0–18.4%, p = 0.0006). Conclusion. A smoking cessation intervention based on best practice guidelines was effective among high risk, pregnant smokers in South Africa.
A number of meta-analyses have concluded that smoking cessation interventions are effective in helping pregnant women quit smoking and in reducing low birthweight and pre-term birth (1–3). However, these conclusions are based on studies conducted in the US, Europe and Australia and it remains unclear whether these achievements can be replicated in developing countries. Whilst tobacco use among women is falling in many Western countries, it is on the increase in the developing world, making tobacco use during pregnancy a growing concern (4).
The ‘5 As’ Smoking Cessation Clinical Practice Guideline (5), which has been specifically adapted for use with pregnant women by the American College of Obstetricians and Gynecologists (ACOG) (6) is currently promoted as best practice for brief cessation counseling by antenatal care providers. The Guideline recommends that a 10–15 minute counseling session by a trained provider and the provision of pregnancy specific, self-help education materials become a standard component of routine antenatal care. Typically, such interventions can double the usual quit rate to about 15% (7).
Antenatal care providers in the public sector in South Africa do not actively and consistently promote smoking cessation among pregnant women and appear to be unaware of best practice guidelines (8, 9). Whilst pregnant smokers are routinely identified, they seldom receive support for cessation beyond very limited, prescriptive advice to quit. Basic educational materials on the risks of smoking are also unavailable.
The purpose of this study was to evaluate the impact of a smoking cessation intervention, which incorporated the ACOG 5As best practice guidelines, on pregnant smokers attending public sector antenatal clinics. To our knowledge, this was the first time that an intervention, based on a model found to be successful in developed countries, was tested in South Africa and, indeed, in a developing country setting.
Material and methods
Study population and setting
The subjects of this study were pregnant women of poor socio-economic status and of mixed ethnic descent living in Cape Town. An alarmingly high percentage (47%) of these women smoke during pregnancy (10).
The study was conducted in four public sector antenatal clinics, which form part of the community based primary health care facilities typically used by low-income communities. They are run by professionally qualified midwives who provide ongoing antenatal care and manage deliveries. A small proportion of pregnant women with high risk pregnancies or complications during delivery are referred to a nearby tertiary level hospital.
The intervention was evaluated using a quasi-experimental design with a natural history cohort and an intervention cohort (Figure 1). The natural history cohort comprised all self-reporting smokers registering at the clinics between February and November, 2006. They served as a control group to determine the quitting rate during pregnancy when receiving usual care. The intervention cohort registered at the same clinics, but a year later, between February and November, 2007, when the intervention was implemented.
The baseline quit rate was assumed to be 7.5% and the quit rate in the intervention group was expected to be 15.2%, based on the findings of a meta-analysis of smoking cessation interventions for pregnant women by Windsor (3).
A two group chi-square test with a 0.05 two-sided significance level has 80% power to detect the difference between a control proportion of 0.080 and an intervention proportion of 0.150 (risk difference of 0.07), when the sample size for each group is 325. For both time periods a drop out rate of 20% was assumed. The sample size needed per time period was determined to be 407. The planned sample size was 450 pregnant women in each year to ensure adequate numbers for analyses. The study was not powered to detect differences in the pregnancy and birth outcomes of the two groups.
All pregnant smokers, identified by using the multiple choice questions recommended by Melvin et al. (11), were invited to participate in the study at their first antenatal clinic visit. The following women were excluded from the study: those who were found not to be pregnant; those beyond 24 weeks gestational age at first booking (calculated on the basis of last menstrual period, palpitation by the midwife or ultrasound); women intending to seek private antenatal care or an abortion; and women who refused to give consent.
Usual care versus the intervention
Pregnant smokers registering at the clinics in 2006 received usual care, which involved midwives asking about smoking status during history taking and then, typically, advising a smoker to quit or reduce smoking in a prescriptive manner. The only complications mentioned were growth retardation and respiratory problems. Earlier researched showed that the midwives' capacity to do counseling was severely limited by time constraints, staff shortages and increased patient load. Midwives also did not have sufficient knowledge to offer women practical assistance in quitting. The only education materials available were general pamphlets on pregnancy with scant information on smoking (9, 12).
Pregnant smokers registering at the same four clinics in 2007 were exposed to the intervention, which was incorporated into routine care. Midwives were trained to use the ACOG 5As Guideline for brief smoking cessation counseling methods. Certain key Motivational Interviewing principles (13) were linked to each step of the guideline, to help midwives adopt a patient centered communication approach in its delivery (Figure 2). Midwives received their training in-service, during two afternoon sessions of two hours each. The training included time for reflection on their current approach, their personal experiences of smoking cessation counseling, as well as opportunities for role play. They were provided with a ‘Health Care Providers Guide to Counselling Pregnant Women about Smoking’, which was adapted from the guide developed by the ACOG (14). Posters summarizing the 5As were put up in the examination rooms to serve as reminders.
Two peer counselors were employed during the intervention to assist the midwives. They were recruited from the local community and had successfully quit smoking themselves. They were trained, along with the midwives, to deliver the 5As. Once the midwife had briefly delivered the first two steps of the Guideline, she informed all smokers about the availability of the peer counselor and explained what assistance she could offer. The peer counselor would repeat steps two and three, tailor the intervention to the pregnant women's stage of change (15) and proceed to steps four and five (Figure 2).
The media produced for the intervention included a self-help Quit Guide, which was adapted, with permission, from Windsor (16). The Guide was tailored to take into account the language, reading levels, cultural values and life circumstances of the target group. In addition, two leaflets, in the format of a popular, local, tabloid newspaper, were produced. The one was tailored for women who were in the pre-contemplation or contemplation phase of behavior change (17), who required further motivation to decide to quit. The other addressed the partners or family members of women who wanted to quit and suggested how best to support her. All the media used photographs and true to life testimonials of pregnant women from the target community in order to facilitate close identification with the materials and their messages. The media was only offered to pregnant women in the context of brief counseling.
The primary outcome measure was quitting, defined as a urinary cotinine level below 100 ng/ml. Secondary outcome measures were: a reduction in smoking, defined as at least half the level of urinary cotinine at study entry, and self-reported quitting, reduction and quit attempts.
Data collection procedures
Women were interviewed by fieldworkers who administered questionnaires during three routinely scheduled clinic visits: (i) at baseline, less than 24 weeks gestation, (ii) mid-pregnancy at 28–35 weeks, and (iii) late pregnancy at 36–39 weeks. Urine samples were collected at the same time points. Written consent was obtained from all participants after a full explanation of study procedures. If women did not attend the routine visits scheduled by the midwives, they were telephoned and arrangements were made to meet them either at home or the clinic for follow up measurements. Patients who were referred to the tertiary hospital for complications were followed up at the hospital.
The baseline questionnaire collected data on socio-demographic characteristics, social support; relationship with the father of the baby, beliefs about smoking and quitting; exposure to environmental tobacco smoke; stage of behavioral change and women's experience of pre-intervention smoking cessation activities at the clinic. Self-reported smoking was measured by number of cigarettes smoked per day and level of addiction determined by the Fagerstrom Test for Nicotine dependence (18). Smoking behavior was re-assessed at times 2 and 3. Alcohol consumption was measured using the definition of a standard drink by Wolmarans et al. (19). Information on the delivery and birthweight was collected from the labor ward register at the clinic or referring hospital. Women in the intervention cohort completed questions which assessed their recall of exposure to the intervention and their level of satisfaction with various intervention components.
Current smoking status at all three measurements was determined by urinary cotinine assay. These were performed by using a Cotinine Direct ELISA kit imported from Bio-Quant Inc. in the USA. The urine samples were taken to the Medical Research Council's laboratory the same day and refrigerated and analysed within 5 days. The laboratory technician was blind to self-reported smoking status and level of tobacco consumption. All women received their results in writing.
To validate the cut off point of 100 ng/ml for active smoking in the study population, cotinine urine tests were conducted on two samples of passive smokers: (1) 123 additional pregnant women recruited from the same clinics, (2) a convenience sample of 62 non-pregnant work colleagues. The cotinine distribution of these two groups of passive smokers was then compared to that of the study population, all of whom were self-reporting active smokers. The definition of 100 ng/ml or above indicating active smoking fitted the smoker distribution well and was maintained for the analysis of the cotinine related study outcomes (Figure 3).
The SAS and STATA statistical programs were used. Frequencies, means and standard deviations were used for the descriptive statistics. Chi square and t-tests were used to test for equivalence between the two cohorts at baseline. Differences in the primary study outcomes were assessed using Chi square and Fischer's exact tests. A 95% confidence interval for the effect size was calculated. The estimation of the intervention effect was based on the intention to treat principle for the primary outcome. The assumption was made that women who were lost to follow up did not quit smoking. Logistic regression models were used to test for an association between quitting and baseline cotinine value and to compare the impact of alcohol consumption on the quitting rates. A Wilcoxon rank sum test was used to test for differences in the median number of quit attempts between the two cohorts.
The Committee for Human Research at the Faculty of Health Sciences of Stellenbosch University in Cape Town granted ethical approval for the study.
Recruitment and follow up
A total of 979 self-reporting pregnant smokers were entered into the study at baseline (443 in the control; 536 in the intervention group). Figure 1 shows the details of the recruitment process, the women excluded after initial recruitment, as well as those women who were lost to follow up by measurements two and three. If women had at least one follow up measurement, they remained in the study.
Description of study participants
There were no significant differences at baseline on demographic variables; with the exception that more women in the intervention cohort had unplanned pregnancies and more of them lived in informal housing (un-serviced, temporary dwellings). Both these variables favored the natural history cohort for better outcomes, rather than the intervention cohort (20).
Twenty-four percent of women in both groups were teenagers between 14 and 19 years old and less than half (NH: 44%; INT: 40%) were pregnant for the first time. Despite over half of women (NH: 54%; INT 56%) not living with the father of the baby, the majority of women said that they were receiving emotional and financial support from them. A significant number also relied on family members, especially their parents for support. More than two thirds of the women reported that they were exposed to environmental tobacco smoke in their homes on a daily basis, indicating a very high prevalence of smoking in this community. The majority of women (NH: 81%; INT: 87%) said that the father of their baby was concerned about her smoking during pregnancy.
There were no significant differences at baseline between the two groups in cotinine profile or self-reported smoking status (Table 1). The mean tobacco consumption according to self-report was around six cigarettes per day, compared to nine per day before becoming pregnant. Level of addiction did not differ significantly either: about half of all women (NH: 45%; INT: 50%; p = 0.110) smoked their first cigarette within 5–30 minutes of waking and could therefore be categorized as heavily addicted to nicotine (18). In both groups, reported and cotinine measured smoking at baseline was positively associated with age, with older women having progressively higher cotinine values.
Table 1. Profile of the two groups at baseline (<24 weeks gestation).
Natural history N = 358
Intervention N = 456
Note: ETS, environmental tobacco smoke.
Mean (standard deviation)
Pregnancy: duration in weeks
Gravidity (no. of pregnancies, including current one)
Parity (no. of viable births)
Age in years
No. of other smokers at home
Housing density (no. of people per room)
> Grade 8
Planned pregnancy (%)
Marital status (%)
Married, living with the father of the baby
Married, not living with father of the baby
Unmarried, living with father of the baby
Unmarried, not living with the father of the baby
Type of house (%)
Formal: brick house/flat
Informal: Wendy house
Bungalow in yard
Shack in squatter camp
Exposure to ETS (%)
Cotinine profile at baseline (%)
Self-reported cigarettes per day Mean (SD)
Other risk behaviors (%)
Used illicit drugs
There were no significant differences in terms of alcohol or drug use at baseline. Binge drinking over weekends was the common pattern. The majority of drug users had used crystal methamphetamine (NH: 78%; INT: 59%) or cannabis (NH: 32%; INT: 48%).
Table 2 summarizes the cotinine validated, point prevalence quit rates for late pregnancy (36–39 weeks gestation). The difference in quitting between the two cohorts was 7.7% (95% CI: 4.6–10.7%, p < 0.0001). As there were hardly any quitters in the Natural History cohort (n = 2), the rate observed in the Intervention cohort was close to the intervention effect. In an intention to treat analysis, where it was assumed that those lost to follow up had not quit, the difference in quitting rates was 5.3% (95% CI: 3.2–7.4%, p < 0.0001).
Table 2. Primary and secondary study outcomes by end of pregnancy (36–39 weeks gestation).
Outcomes frequency (%)
Natural history cohort (N = 269)
Intervention cohort (N = 358)
Risk difference between groups
Primary outcome: quitting by urinary cotinine (<100 ng/ml)
(A) Analysis excluding lost to follow up (NH: N = 72) (INT: N = 60)
(B) Intention to treat analysis (those lost to follow up, classified smokers)
Quitting by self-report
Reduction by urinary cotinine (value reduced by at least 50%)
Reduction by self-report
Quit attempts by self-report Mean (SD)
A logistic regression model for the intervention cohort showed that there was no significant association with baseline cotinine levels. Quitters came from all categories of smokers, including those smokers with high baseline cotinine readings.
Table 2 shows that by end of pregnancy, 101 (28%) of women in the intervention cohort had reduced their cotinine levels by at least half, compared to 44 (16%) in the natural history cohort - a difference of 11.8% (95% CI: 5.0–18.4%, p = 0.0006).
Self-reporting quitting rates also differed significantly in late pregnancy. The data in Table 2 shows a large discrepancy with cotinine validated quitting and self-reports. Only 30 of the 66 (45%) women in the intervention group and 2 of the 37 (5%) women in the natural history group had cotinine readings commensurate with their self-reported quit status.
Women in the intervention cohort reported significantly more quit attempts (defined as not having even a puff for at least 24 hours) than women in the natural history cohort. This indicates a move to the stage of preparation for change, which is characterized by experiments in adopting the new behavior (17).
Although the study was not powered to detect a significant difference in any pregnancy outcomes, a comparison of birthweight between the two cohorts showed a difference of 44 g, which was not statistically significant. The mean birthweight was 2949.5 g in the intervention cohort and 2905.7 g in the natural history cohort.
There was a notable drop in self-reported alcohol consumption from baseline to end of pregnancy, but this trend occurred in both cohorts and was not significantly different (NH: 54–19%; INT: 49–14%). When modeling quitting on alcohol status at baseline, the quitting rate was lower in the sub-group reporting alcohol use, but the differential was not statistically significant. The estimated intervention effect was 9.1% (95% CI: 4.5–13.7%) in the non-drinkers compared to 6.0% (95% CI: 2.1–9.9%) in the drinkers.
As can be seen in Table 3, dramatically fewer women in the natural history cohort than the intervention cohort reported they had been offered educational material on smoking or that a midwife had discussed smoking with them since coming to the clinic. Table 4 presents further data relating to the implementation of the intervention. The midwives referred women to the peer counselors with great alacrity and 97% of women reported choosing to see the peer counselor. Midwives were also reasonably efficient in delivering the first step of the 5As, as is shown in the marked improvement in their asking women about smoking between 2006 and 2007. As was intended, it was mainly the peer counselors who delivered the remainder of the 5A Guideline.
Table 3. Comparison of smoking cessation activities in clinic prior to and during intervention.
Mid-pregnancy measurement (28–35 weeks gestation)
End-pregnancy measurement (36–39 weeks gestation)
Did the nurse/midwife ask/talk to you about smoking?
Natural history cohort
N = 350
N = 269
Yes: frequency (%)
N = 444
N = 375
Yes: frequency (%)
During your visits to the clinics, were you offered any education material on smoking?
Natural history cohort
N = 350
N = 269
Yes: frequency (%)
N = 444
N = 375
Yes: frequency (%)
Table 4. Implementation of intervention.
Intervention cohortProcess questions: End of pregnancy measurement (36–39 weeks gestation)
N = 375
(1) Referral to peer counselor: frequency (%)
During your visits to the clinic, were you informed about the Stop Smoking counselor?
If yes, did you go and see her?
(2) Brief counseling: frequency (%)
Did anyone at the clinic ask if you were willing to try and quit? (Step no. 3 in 5A protocol: ASSESS stage of change). If so, who was it?
Stop Smoking counselor
Who informed you that smoking would be discussed again during your next visit? (Step no. 5 in 5A protocol: ARRANGE for follow up)
Stop Smoking counselor
(3) Provision of education materials: frequency (%)
Were you offered any education material? (Step no. 4 in 5A protocol: ASSIST in decision and strategies for quitting)
Yes, quit newspaper
Yes, newspaper for family/partner
Yes, self-help quit guide
Other, e.g. pamphlet
Who offered you the material?
Stop smoking counselor
(4) Use of materials: frequency (%)
Did not take any materials
% who read quit newspaper
% who read newspaper for family/partner
% who read self-help quit guide
(5) Did anyone else read the materials?
Father of baby
Other, e.g. boyfriend's mother
Rating of intervention components: frequency (%)
(1) Do you agree with the following statement? The midwife/peer counselor was respectful when she spoke to me about smoking
They did not discuss smoking with me
(2) What did you think of her counseling about smoking?
Did not receive counseling
(3) What did you think of the education material?
Nothing, because I was not given any material
I did not read it
Table 4 illustrates that although women were positive about the attitudes of the midwives, they agreed more strongly that the peer counselors were respectful when talking to them about smoking. Similarly, about twice as many women rated the peer counselors and the education materials as ‘very helpful’ in comparison to the midwives.
This study demonstrated that the 5As best practice guideline for smoking cessation interventions for pregnant women was effective in a developing country setting. The cessation rate improvement in the intervention Group of 7.7% (or 5.3% in the intention to treat analysis) was modest, but coincides with the average increase of 6.4% achieved by the studies synthesized in the latest Cochrane review of smoking cessations interventions during pregnancy (2), most of which were undertaken in high-income countries. It was encouraging that our intervention succeeded in assisting even heavy smokers in quitting, as the 5As model has tended to be less effective with this group (7). The level of cotinine validated smoking reduction in the intervention cohort was also a positive outcome and supports the view that harm reduction, as well as quitting be given due attention in such interventions (21). Although the study was not powered to detect differences in any pregnancy outcomes, the 44 g increase in birthweight in the intervention cohort is worth noting and is in line with the findings of other trials (2).
The low rate of quitting (under 1%) among women in the natural history cohort receiving usual care is indicative of the public antenatal service's current failure to address the issue of smoking and underscores the need to create a policy and practice environment in South Africa favorable to the adoption of such interventions. It remains for further research to determine what additional elements of intervention could produce an increase in effect size. Providing incentives appears to be a promising strategy (2) and the use of more intensive behavioral therapy, combined with judicious use of nicotine replacement therapy (NRT) with heavy smokers may need to be considered once it is clear that brief counseling has failed. However, the safety of NRT during pregnancy is yet to be clearly established (6) and the cost may make it impractical in a resource poor setting.
The results of the process evaluation indicate that the intervention was successfully delivered and met the main criteria proposed by Walsh and Redman (22) for assessing whether a smoking cessation intervention is suitable for routine application within antenatal care services: the program elements were well received by pregnant women and midwives, time commitments for training and delivery of the intervention were feasible, the intervention was easily incorporated into usual antenatal clinic procedures and the education materials were readily available.
Public antenatal clinics in South Africa represent a particular challenge for the dissemination of interventions because of acute staff shortages, high work load and the complexity of meeting the needs of growing numbers of pregnant patients with HIV/AIDS, TB and chronic disease. The extent to which the midwives in this study relied on the peer counselors to implement the intervention is an indication of these constraints on their time. However, their support for the intervention and their appreciation of the peer counselor's presence were unequivocal, as was demonstrated by their 100% referral of women to them. Women's openness in seeking the support offered by the peer counselors (97% choose to see them), as well as their high ratings of their helpfulness is an indication of their need for an opportunity to discuss their concerns about their pregnancy and their positive predisposition to such social support. The success of peer counseling, particularly with low-income women, appears to be related to the non-hierarchical relationship created through the sharing of life experiences, by increasing social support and decreasing stress (23). Whilst peer or lay counseling has shown success in other aspects of prenatal care, such as breastfeeding (24), it has not been widely tested for smoking cessation and the few reported studies show mixed results (25, 26). However, the potential role of such lay counselors in behavioral change counseling deserves further investigation in the South African context, as they may well be a cost effective means of addressing the problem of acute staff shortages that is highly acceptable to both midwives and pregnant women.
The finding that 52% of pregnant women had drunk alcohol since becoming pregnant and 14% had used drugs is of grave concern. The concurrence of other risk behaviors and the failure to specifically address these may have limited the success of this smoking cessation intervention. The experience of the peer counselors in this study showed that for many pregnant women, smoking was a useful entry point to discuss other more stigmatized behaviors, which were of greater concern to them than their nicotine addiction. Behavioral counseling interventions with such populations may need to be more responsive to addressing the multiple and often associated, risk behaviors of pregnant women and avoid focusing on a single health issue (27). The model of intervention and training used in this study could be easily adapted to accommodate the various counseling needs of pregnant women, as long as the necessary supporting educational materials are available and an effective referral system for more intensive counseling by trained professionals is put in place.
A randomized control trial would have been the optimal research design, but this was not feasible due to the costs involved and because there are insufficient number of clinics serving this particular study population in Cape Town.
The loss to follow up was greater than anticipated. Despite this, the intention to treat analysis showed an intervention effect. The main reason for loss to follow up, in both years, was delivery before second measurement. This was mainly due to errors in the calculation of gestational age. As ultrasound is not available at the clinics, midwives rely on palpitation and the date of the last menstrual period to calculate gestational age, both of which are often inaccurate. Genuine premature delivery would also have accounted for some of these cases. A further important reason for loss to follow up was the difficulty in tracing women who did not attend the clinic for their regular scheduled visits. Efforts were made to contact these women, but many of them did not have phones or were not living at the address they had provided. Home visits were done where possible. The high rates of attrition and the reasons for them in our study are common in trials involving pregnant women of poor socio-economic status (2).
It is well known that the rates of relapse among pregnant women are high (17). A limitation of this study is that we did not have the resources to follow up women post-delivery and the relapse rates among the 32 quitters remains unknown.
Smoking during pregnancy is an important, preventable risk factor for the most common causes of intrauterine and early neonatal death among disadvantaged women of mixed ethnic descent in South Africa (28). As illustrated by this study, a cessation intervention which used trained peer counselors to deliver an adapted form of the ACOG Guideline and provide tailored educational and self-help materials was successfully implemented in public sector, antenatal clinics serving this high risk population.
Future studies need to investigate the addition of further components which may enhance the modest quit rates achieved and the possibility of applying this intervention model to address multiple risk behaviors in pregnancy.
This work was carried out with the aid of a grant from the Research for International Tobacco Control (RITC), an international secretariat housed at the International Development Research Centre (IDRC) in Ottawa, Canada.
We are very grateful to the midwives of Bishop Lavis, Elsies River and Tygerberg for their support and to the Department of Health for granting us access to the clinics.
Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.