|Home | About | Journals | Submit | Contact Us | Français|
In studies of the fetal origins of disease and life course epidemiology, measures of fetal exposure may be based on information reported by the adults who were exposed in utero. In particular, the full spectrum of consequences of in utero exposure to maternal tobacco smoking is now an area of active investigation, and the ability to report such exposure reproducibly is of interest. We evaluated the reproducibility of in utero exposure to tobacco smoke, reported by the adult daughter during consecutive pregnancies.
This study was based on 11,257 women who enrolled for more than one pregnancy in the Norwegian Mother and Child Cohort Study (MoBa). Participants completed a questionnaire around 17 weeks of gestation, which asked about their in utero exposure to tobacco smoke. Kappa statistics were calculated. Determinants of agreement were evaluated using logistic regression.
Weighted Kappa for in utero exposure for the first and second reports was 0.80. Determinants of agreement were higher education (better) and longer time between reports (worse).
Information on in utero exposure to maternal tobacco smoking provided by adult women was highly reproducible in this population.
About 10–30% of women in western countries smoke during pregnancy (1, 2). The short-term risks of maternal smoking to offspring, such as being delivered preterm or small-for-gestational-age, have been well-characterized (3, 4). Longer-term consequences of such in utero exposure, however, are still under investigation. For example, such exposure may increase risk of early menarche, infertility, premature menopause, obesity, type 2 diabetes, and increased serum cholesterol (5–10). Investigations of longer-term consequences often depend on adults’ reports of their exposure in utero (11). Information about the reproducibility of adults’ reports will enhance the interpretation of data on health effects. In a large pregnancy cohort in Norway, women could enroll for more than one pregnancy, thus providing data on their in utero tobacco exposure more than once. We used these data to study the reproducibility of reporting.
This study is based on the Norwegian Mother and Child Cohort Study (MoBa), conducted by the Norwegian Institute of Public Health (12). Enrollment was from 1999 to 2008 and included about 107,000 pregnancies from 90,000 women. During enrollment, the proportion of pregnant women in Norway invited to participate gradually increased, to about 71%. This proportion never reached 100% because two large hospitals in Oslo did not enroll subjects, and because women were recruited only through hospitals and maternity units with at least 100 deliveries per year. The response rate was about 44% (13). Pregnant women were recruited with a mailed invitation before a routine ultrasound examination offered to all pregnant women in Norway at 17–18 weeks of gestation (www.fhi.no/morogbarn). Other than noted above there were no exclusion criteria (12). The Regional Committee for Medical Research and the Norwegian Data Inspectorate approved the study and informed consent was obtained from each participant. The present study is based on participants recruited from 1999 to 2007 (MoBa database version 4.201). Participants were asked to complete a questionnaire about demographic characteristics, reproductive health, disease and medication history, lifestyle, and socioeconomic status.
Women were asked: “Did your mother smoke when she was pregnant with you?” Those who answered “yes” to this question were classified as having been exposed to tobacco smoking while in utero; those who responded “no” were considered unexposed. Women who answered “don’t know” or who did not respond to this question were classified as unknown exposure.
Among the participants born in 1967 or later (when the Medical Birth Registry of Norway was created [n=11,082]), birth weight was available for 90.8%, and was used as a surrogate to validate their reported in utero exposure to tobacco smoke.
Among women with more than one pregnancy in MoBa, 11,257 completed the first study questionnaire twice. To evaluate the pairwise agreement between the two reports, we calculated the kappa statistic (κ), the fraction of the observed agreement not due to chance relative to the maximum non-chance agreement (14). Weighted kappa (κw) (15) was also calculated for the 3 × 3 tables using the reported in utero exposure to maternal smoking as an ordinal variable as suggested by Maclure & Willett (16) with the following categories: yes, unknown and no. In this strategy, non-contiguous categories are considered to have worse disagreement than contiguous categories. Standard linear weights were used (17).
We also examined determinants of agreement between responses among women who had completed the questionnaire twice. For this purpose we used the logistic model of agreement where the outcome variable is equal one when the two reports agree (cells in the diagonal of 2 × 2 or 3 × 3 tables) and zero if not (18). In a separate analysis, we used logistic regression to assess the determinants of the “unknown” status for in utero exposure to maternal smoking. The outcome variable was coded as one if the first or second report had an answer of “unknown” status for in utero exposure, and zero otherwise.
The mean age of the study participants was 28.1 years (standard deviation, 3.9) during the reporting period. Women who reported being exposed to tobacco smoke in utero were lighter at birth compared to those not exposed; this was observed among all births and among those born after 37 completed weeks of gestation. Among women whose in utero exposure was unknown, the mean birth weight were intermediate for all births and term births (Table 1). The adjusted birth weight difference (non exposed-exposed, adjusted for gestational age, age and parity of the participant’s mother) was 178 g for all births and 181 g for term births (data not shown). The overall weighted Kappa for the 3 × 3 table (yes, unknown, no) was 0.80 for the first and second reports of in utero exposure to maternal smoking (Table 1). When the analysis was restricted to those whose responses were either yes or no at both times (2 × 2 table), the kappa statistic was 0.93.
Results from the logistic model based on 2 × 2 and 3 × 3 tables are presented in Table 2. More educated participants (College and > College) were significantly more likely to report the same exposure at both times as compared to those with less than High School. A longer time interval between reports (4 – 8 years) was significantly associated with poorer agreement. An interval of 3 years between reports was also a significant predictor of poorer recall in the 3 × 3 analysis.
In the analysis of predictors of “unknown” in utero exposure status (data not shown), we observed that women with higher education were less likely to report “unknown” status of exposure (College, OR = 0.60, 95%CI: 0.48 – 0.74; > College, OR = 0.62, 95%CI: 0.49 – 0.78) as compared with those with less than a high school education. Those with longer time between reports were more likely to report “unknown” status of exposure (4 – 8 years, OR = 1.21, 95%CI: 1.01 – 1.44). Woman’s age and current smoking status were not significantly associated with “unknown” in utero exposure status.
In the present study the reproducibility of daughters’ reports on in utero smoking exposure was much higher than expected by chance, and consistent with a previous study in the U.S. (19). An offspring’s report about their mother’s smoking status during pregnancy has been previously shown to be reasonably valid (Table 3) (11, 20, 21). In the present study the reproducibility was affected by whether subjects with unknown exposure to in utero tobacco smoking were included in the calculation of κ. When the effect of this inclusion is considered in interpretation of Table 3, the data suggest that the kappa for the reproducibility of reporting this exposure is higher than for the validity. For example, when we excluded the unknown category from the calculations, as was done in the validity studies, we observed a higher coefficient (κ=0.93).
In one of the studies of agreement between reports from mothers (about her own smoking during pregnancy) and daughters (about her in utero exposure), the concordance was lower among older daughters (11). In the present study, participants’ age was not a significant predictor of agreement and weighted kappas were comparable across categories of age (Table 2), similar to findings in two previous studies (19, 21). Our results showed that college education (and > college) was consistently associated with better agreement, which was not observed in a previous study (21). However, in the study by Sandler and Shore (21) only 28 % had more than high school education, while in our study the corresponding percentage was much higher (69.2 %). Although our study was limited to women, similar agreement was observed by gender in the two previous studies that also included males (19, 21). As expected, longer time between reports was associated with poorer reproducibility in this study. However, kappas remained high across categories of time between reports even when the unknown category was included in the calculations. Consistent with this result, we observed that longer time between reports was a significant predictor of unknown status of exposure.
Our study had several limitations. Women who participated in MoBa were not a representative sample of all pregnant women from Norway during the period of interest, and previous analyses suggest that prevalence of exposures and outcomes are biased due to self-selection (e.g. smokers were under-represented and multivitamin users over-represented) (13). An underestimation of the prevalence of in utero exposure to maternal smoking was expected, because women who reported being smokers were more likely to report in utero exposure to maternal smoking (data not shown). In Norway, during the period when the study participants were born (between 1960 and 1989), the prevalence of smoking among reproductive age females (16 to 55 years) ranged from 26% to 52%. The prevalence of daily smokers among women ages 25 to 34 years was 46% in 1974 and 41% in 1977 (22, 23). In the present study, 38% of the participants in were born between 1974 and 1977; and the prevalence of reported in utero smoking was 23% for those born in 1974 and 26% for those born in 1977. It is also likely that women who participated in MoBa for more than one pregnancy may have been especially health conscious, and therefore may have been better-than-average reporters of early life exposure.
Another limitation is that we did not have data on the number of cigarettes smoked per day by the subject’s mother. Previous studies that evaluated the reproducibility or validity of amount of cigarettes found moderate to poor agreement (19, 21). In another study, however, agreement was better when the amount of cigarettes smoked by the mother was greater (15 + cig/day) (11).
Although we were not able to evaluate the validity of in utero smoking directly, we were able to show that this exposure can be regarded as valid based on the birth weights of a subset of the MoBa participants. Maternal smoking during pregnancy has been associated with an average reduction of 149 g in birth weight (24), and in the present study we observed a 181 g average decrement.
In summary, reports of in utero exposure to maternal tobacco smoking have consistently been shown to be highly reproducible, and the degree of reproducibility depends on whether subjects with unknown exposure were included in the analysis.
This study was supported in part by the Intramural Research Program, the National Institute of Environmental Health Sciences (NIEHS), the National Institutes of Health (NIH). The Norwegian Mother and Child Cohort Study is supported by the Norwegian Ministry of Health, NIH/NIEHS (grant no N01-ES-85433), NIH/NINDS (grant no.1 UO1 NS 047537-01), and the Norwegian Research Council/FUGE (grant no. 151918/S10).
CONFLICT OF INTEREST: None declared
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.