Self-reported nicotine exposure and plasma levels of cotinine in early and late pregnancy
Abstract
Background. The accuracy of maternal self-reported smoking information, especially exposure to environmental tobacco smoke, has been questioned. This study aimed to validate self-reported smoking, smoking cessation, and environmental tobacco smoke exposure in early and late pregnancy, using the biomarker cotinine as the gold standard. Methods. Prospective cohort study of 953 pregnant Swedish women between 1996 and 1998. In-person interviews and cotinine measurements were performed at 6–12 and 31–34 completed weeks of gestation. All women were asked about nicotine exposures throughout pregnancy, including cigarette smoking, oral snuff, nicotine replacement therapy, and environmental tobacco smoke exposure. Results. The validity of self-reported daily smoking was high in early and late pregnancy. However, among women reporting smoking cessation before the first interview and between the first and second interviews, 13% and 25% misreported active smoking, respectively. According to cotinine measurements, 22% of non-smoking women were exposed to environmental tobacco smoke in early pregnancy, and 8% were exposed in late pregnancy. Self-reported information on environmental tobacco smoke exposure in early and late pregnancy misclassified most exposed women as unexposed. Conclusion. The results of this study indicate that self-reported smoking information among pregnant women can be trusted. However, among women reporting smoking cessation during pregnancy, the misclassification rate increased with recency of quitting. Environmental tobacco smoke exposure was common among non-smokers, and the low validity of self-reported environmental tobacco smoke exposure suggests that future studies on environmental tobacco smoke exposure and risks of pregnancy outcomes may have to rely on biomarker assessments.
Abbreviations
-
- ETS
-
- environmental tobacco smoke
-
- CI
-
- confidence interval
-
- PPV
-
- positive predictive value
-
- NPV
-
- negative predictive value
-
- LR
-
- likelihood ratio
Studies on maternal smoking and adverse pregnancy outcomes are generally based on self-reported exposure information. A Swedish study by Lindqvist et al. (1) biochemically validated self-reported smoking in the first trimester, and found that 6% of reported non-smokers had cotinine values corresponding to active smoking. However, many women state that they quit smoking either when they find out that they are pregnant, or later during pregnancy (2). It is not clear whether misclassification of smoking exposure is influenced by recency of smoking cessation.
Prenatal exposure to environmental tobacco smoke (ETS) has been associated with risk of fetal growth restriction and preterm birth, and possibly with spontaneous abortion (3, 4). Studies of the effects of prenatal ETS exposure rely to a great extent on maternal self-report. It is not certain that self-reported information can be used to study ETS exposure, as total ETS exposure includes exposures at various places, i.e. at home, at work, and in public places (5).
The aims of this prospective study were to determine the accuracy of maternal self-reports of active smoking, smoking cessation during pregnancy, and ETS exposure during pregnancy, using cotinine measurements as the gold standard.
Material and methods
From January 1996 through December 1998, 1,037 women were asked to participate in a prospective cohort study. Criteria for participation were that the women were Swedish-speaking, living and seeking prenatal care in Uppsala County, and pregnant in the first trimester when entering the study. The participation rate was 92% (n = 953). In-person interviews and cotinine measurements were conducted twice during pregnancy: in early and in late pregnancy (6–12 and 31–34 completed weeks of gestation, respectively). Further methodological details have been reported previously (6).
Nicotine exposure information
Three specially trained midwives conducted in-person interviews with all participants using a structured questionnaire. In early pregnancy (6–12 weeks of gestation), women were asked about all nicotine exposures, including active smoking, exposure to environmental tobacco smoke (ETS), use of oral snuff, transdermal nicotine patches, and nicotine chewing gum. This information was collected on a weekly basis, starting four weeks before last menstrual period until the most recently completed week of gestation. Information on ETS exposure was recorded as daily exposed (numbers of hours per day), non-daily exposed, and non-exposed.
In late pregnancy (31–34 weeks of gestation), information on nicotine exposures was asked on a biweekly basis from the time of the first interview until the last completed week of gestation. Nicotine exposure information was categorized in the same way as in the first interview.
Cotinine analysis
Women were asked to provide blood samples at the time they were interviewed in both early and late pregnancy. Blood samples were kept frozen at −80°C until assayed. Plasma cotinine was measured by gas chromatography with use of N-ethylnorcotinine as an internal standard (detection limit 0.1 ng/ml) (7). A plasma cotinine level of more than 15.0 ng/ml was defined as indicating active smoking, and between 0.1 and 15.0 ng/ml among non-smokers was defined as indicating exposure to ETS (8).
Study population
Of the 953 women completing the first interview, 885 women also completed the second interview, of whom 52 women had missing cotinine measurements in early or late pregnancy (Figure 1). Further, 48 women reported daily use of other nicotine containing products (i.e. oral snuff or nicotine replacement therapy) in early or late pregnancy. A majority of these women had cotinine levels > 0.1 ng/ml. As these women were likely to have detectable cotinine levels not caused by active smoking or ETS exposure (9), they were excluded from the analyses.
Flow chart of study participants.
*6–12 weeks of gestation.
†22 women miscarried, 6 electively terminated their pregnancy, 16 delivered before the second interview, 12 opted to withdraw from the study, 6 moved outside Uppsala County, and 6 were lost to follow-up.
‡31–34 weeks of gestation.
§Self-reported smoking information in early pregnancy.
Statistical analysis
We estimated the validity of self-reported smoking, smoking cessation during pregnancy, and self-reported ETS exposure in early and late pregnancy, using cotinine measurements as the gold standard. Sensitivity, specificity, positive predictive values (PPV), and negative predictive values (NPV) were calculated and 95% confidence intervals (CI) computed by the exact method based on the binomial distribution. The likelihood ratio (LR) for a positive test was algebraically calculated as the sensitivity/(1 − specificity), and the LR for a negative test was calculated as (1 − sensitivity)/specificity (10).
Oral informed consent was obtained from all women. The ethics committee of the medical faculty at Uppsala University approved the study.
Results
According to cotinine measurements, 8% (n = 66) of the women in the cohort were smokers (cotinine > 15.0 ng/ml) in early pregnancy (6–12 weeks of gestation), and 7% (n = 58) were smokers in late (31–34 weeks) pregnancy (Table I). In early pregnancy, median and mean plasma cotinine levels among cotinine-indicated smokers were 91.0 ng/ml (90% central range 21.5–228.1 ng/ml) and 103.1 ng/ml (SD 71.3 ng/ml), respectively. In late pregnancy, corresponding values were 90.1 ng/ml (17.8–265.3 ng/ml) and 99.5 ng/ml (SD 69.4 ng/ml), respectively.
| Cotinine (ng/ml) | ||||
|---|---|---|---|---|
| Daily smoking | >15.0 | ≤15.0 | Total | Likelihood ratio |
| Early pregnancy* | ||||
| Yes | 53 | 7 | 60 | 82.5 |
| No | 13 | 712 | 725 | 0.2 |
| Total | 66 | 719 | 785 | |
| Late pregnancy† | ||||
| Yes | 45 | 3 | 48 | 188.0 |
| No | 13 | 724 | 737 | 0.2 |
| Total | 58 | 727 | 785 | |
- *6–12 weeks of gestation. Sensitivity 80% (95% CI 69–89); specificity 99% (95% CI 98–99.6); positive predictive value 88% (95% CI 77–95), negative predictive value 98% (95% CI 97–99).
- †31–34 weeks of gestation. Sensitivity 78% (95% CI 65–87); specificity 99.6% (95% CI 98.8–99.9); positive predictive value 94% (95% CI 83–99); negative predictive value 98% (95% CI 97–99).
The sensitivity and specificity of self-reported active smoking were high both in early and in late pregnancy (Table I). Among 66 women in early pregnancy with cotinine levels >15.0 ng/ml, 53 reported daily smoking, providing a sensitivity of 80%; whereas among 719 women with cotinine levels ≤ 15.0 ng/ml, 712 reported non-smoking, providing a specificity of 99%. PPV and NPV of self-reports were also high throughout pregnancy. In early as well as in late pregnancy NPV was 98%, meaning that among self-reported non-smokers, only 2% under-reported active smoking. The LR also indicate that self-reported smoking information is useful and can be trusted.
Next, we validated self-reports of smoking habits during pregnancy among the 122 women who stated that they were smokers at the time of conception. Using cotinine measurements as the gold standard provided fairly high sensitivity and specificity (87% and 89% respectively; Table II). Among 62 women who reported smoking cessation before the first interview, 54 women had cotinine values ≤ 15.0 ng/ml, giving a NPV of 87%. Thus, among self-reported quitters between conception and the first interview, 13% (8 out of 62) misreported active smoking.
| Cotinine (ng/ml) | ||||
|---|---|---|---|---|
| >15.0 | ≤15.0 | Total | Likelihood ratio | |
| Smokers at conception (n = 122) | ||||
| Daily smoking, early pregnancy* | ||||
| Yes | 53 | 7 | 60 | 7.6 |
| No | 8 | 54 | 62 | 0.1 |
| Total | 61 | 61 | 122 | |
| Smokers in early pregnancy (n = 60) | ||||
| Daily smoking, late pregnancy† | ||||
| Yes | 42 | 2 | 44 | 6.4 |
| No | 4 | 12 | 16 | 0.1 |
| Total | 46 | 14 | 60 | |
- *6–12 weeks of gestation. Sensitivity 87% (95% CI 76–94); specificity 89% (95% CI 78–95); positive predictive value 88% (95% CI 77–95); negative predictive value 87% (95% CI 76–94).
- †31–34 weeks of gestation. Sensitivity 91% (95% CI 79–98); specificity 86% (95% CI 57–98); positive predictive value 95% (95% CI 85–99); negative predictive value 75% (95% CI 48–93).
Among 60 women who stated that they were daily smokers in early pregnancy, 44 women also stated that they were smokers in late pregnancy (Table II). Among the 60 smokers in early pregnancy, self-reported smoking in late pregnancy was compared to cotinine, providing a sensitivity of 91% and a specificity of 86%. Among the 16 women reporting smoking cessation between the first and second interview, 12 had cotinine ≤ 15.0 ng/ml, giving a NPV of 75%. Thus, among self-reported quitters between the first and second interview, 25% (4 out of 16) misreported active smoking in late pregnancy.
Finally, we validated self-reported exposure to ETS against cotinine measurements among non-smokers (defined as women reporting no smoking in the last week before the interview and who had a cotinine level ≤ 15.0 ng/ml). Cotinine measurements were dichotomized into ETS-exposed (0.1–15.0 ng/ml) and non-exposed (<0.1 ng/ml). Self-reported ETS exposure was dichotomized into daily (≥1 h/day) and non-daily exposed. According to cotinine measurements, 22% (n = 159) of non-smoking women were exposed to ETS in early pregnancy, and 8% (n = 55) were exposed in late pregnancy (Table III). In early pregnancy, median and mean plasma cotinine levels among cotinine-indicated ETS-exposed women were 3.4 ng/ml (90% central range 0.6–8.9 ng/ml), and 3.8 ng/ml (SD 2.3 ng/ml), respectively. In late pregnancy, corresponding values were 1.9 ng/ml (0.7–7.7 ng/ml) and 2.4 ng/ml (SD 1.8 ng/ml), respectively. Among 159 cotinine-indicated ETS-exposed women in early pregnancy, only 9 reported daily exposure (sensitivity 6%), while among 32 women stating exposure to ETS in early pregnancy, 9 were exposed according to cotinine measurements (PPV 28%, Table III). Thus, self-reported information on ETS exposure misclassified most exposed women as unexposed. Similar results were also obtained in late pregnancy. The low LR of a positive test in early (LR 1.4) and in late (LR 1.1) pregnancy also indicate that asking pregnant women about ETS exposure is not very useful.
| Cotinine (ng/ml) | ||||
|---|---|---|---|---|
| Daily ETS exposure* | 0.1–15.0 | <0.1 | Total | Likelihood ratio |
| Early pregnancy (n = 712)† | ||||
| Yes | 9 | 23 | 32 | 1.4 |
| No | 150 | 530 | 680 | 1.0 |
| Total | 159 | 553 | 712 | |
| Late pregnancy (n = 724)‡ | ||||
| Yes | 2 | 22 | 24 | 1.1 |
| No | 53 | 647 | 700 | 1.0 |
| Total | 55 | 669 | 724 | |
- *Daily ETS exposure was classified as ≥ 1 h/day, and non-daily exposure as < 1 h/day.
- †6–12 weeks of gestation. Sensitivity 6% (95% CI 3–10); specificity 96% (95% CI 94–97); positive predictive value 28% (95% CI 14–47); negative predictive value 78% (95% CI 75–81).
- ‡31–34 weeks of gestation. Sensitivity 4% (95% CI 0.4–13); specificity 97% (95% CI 95–98); positive predictive value 8% (95% CI 1–27); negative predictive value 92% (95% CI 90–94).
Discussion
In this study cotinine levels were measured repeatedly during pregnancy for validation of self-reported data on active smoking, smoking cessation, and ETS exposure. We found that self-reported smoking information had a high validity, but misclassification increased with recency of smoking cessation. ETS exposure, as indicated by cotinine levels, was common among non-smokers, but the validity of self-reported ETS exposure information was poor.
In the present study self-reported smoking information had a high validity in early as well as in late pregnancy. Our finding in early pregnancy is consistent with another Swedish study (1), while other validation studies have found self-reported smoking less reliable (8, 11), (12). In this cohort, 8% of women were self-reported active smokers at the first interview, while the overall rate of self-reported smoking at registration to prenatal care in Uppsala County was 12% during the study period (13). We cannot rule out that our cohort represents a selected group of more healthy individuals than the general population.
Fifteen percent of all women in this cohort reported smoking at the time of conception. When we only evaluated self-reports within this group, we found that many women under-reported active smoking in early pregnancy. Moreover, under-reporting of active smoking was even more common in late pregnancy among women who reported smoking in early pregnancy. An implication from this finding would be that anti-smoking counseling during prenatal care could be intensified for women reporting smoking cessation before or during pregnancy.
A considerable proportion (22%) of non-smoking pregnant women in this cohort was exposed to ETS according to cotinine levels. Although this figure may seem surprisingly high, the rate is in agreement with other studies on pregnant women using cotinine measurements to define ETS exposure (8), (14–16). The association between maternal self-reported ETS exposure information and cotinine measurements has been investigated in a limited number of studies, which have found poor (17) to moderate (18–20) correlations. We found that self-reported information on ETS exposure misclassified most exposed women as unexposed. In our study, self-reported information was based on number of hours exposed, but did not disentangle different possible sources of exposures, i.e. ETS exposure at work, at home, and in public places. It is possible that the precision of self-reports could have been improved with a more extensive interview. However, the magnitude of ETS exposure is influenced by duration of exposure and the air concentration of ETS (determined by number of smoked cigarettes, smoking pattern, distance to smokers, room size, furnishing, and ventilation) (4), which may be problematic to recall or quantify for exposed persons. We believe that the findings in the present study support the use of biomarkers for exposure assessments in future studies on maternal ETS exposure and risks of adverse pregnancy outcomes.
Cotinine is a widely used biomarker of active smoking and of ETS exposure (8, 21). The half-life of cotinine is approximately 17 h for non-pregnant women (21), and close to 9 h among women between 16 and 40 weeks of gestation (22). Cotinine is a nicotine-specific metabolite, and detectable levels can, besides tobacco smoke, also be caused by other nicotine products including oral snuff and nicotine replacement therapy (9). In Sweden, the use of oral snuff among young women has become increasingly common (13). In order to avoid misclassification of exposure status, we therefore excluded women who used oral snuff or nicotine replacement therapy from the study population.
The accuracy of smoking information may be inflated in validation studies when women are aware of biomarker measurements (23). When this study was planned women were informed about the general aims of the study, and that blood samples were analyzed. However, women were not specifically informed that smoking habits were studied, or that biomarkers of nicotine were measured in their blood samples. A structured questionnaire was used at the interviews, and many other risk factors were considered; thus, the interview situation did not focus solely on tobacco use. From an ethical and informed consent point of view this procedure might not have been approved today. Still, we believe this is a scientific strength, as the results may be less inflated by intervention bias.
There are at least three limitations to consider in this study. First, we lacked information on occasional smoking. There is a possibility that women smoking cigarettes on an intermittent basis might be misclassified as exposed to ETS. Second, there is a delay of 1–6 days (mean 3 days) between self-reported data and cotinine measurements, and women who changed their smoking habits during this interval may have been misclassified. Third, the choice of cut-off level of cotinine differs between previous validation studies. We used the cut-off level 15.0 ng/ml to distinguish active smokers from non-smokers, whereas others have used a lower cut-off (19, 24). Concordant with previous reports (4), the range for cotinine values among smokers and ETS-exposed was well separated also in this study, and using the cut-off 10 ng/ml in additional analyses did not change our main findings (data available at request).
To summarize, we found that self-reported smoking information in this prospective study had a high validity. However, as self-reported quitters during pregnancy were more likely to under-report active smoking, we believe that all women who report smoking around time of conception would benefit from prenatal smoking intervention. The high prevalence of ETS exposure among non-smoking pregnant women is disturbing and may have substantial public health consequences. As self-reports were an imprecise measure of ETS exposure in this as well as in previous studies, forthcoming exposure assessment may have to rely on biomarker measurements.
Acknowledgement
Supported by the International Epidemiology Institute through a grant from the National Soft Drink Association.
