Discussion
Our findings indicate that prenatal SHS exposure increases the risk of late-onset AD, especially in sensitized school-age children. While the relationship between maternal urine cotinine levels and AD in school-age children could not be explored, we noted a relationship between higher maternal urine cotinine levels and the risk of AD symptoms in preschool children (ages 4–6). These results provide strong scientific support for our observations. Our analyses of the AD phenotypes have reported the effects of prenatal SHS exposure on late-onset AD. The present study implies that children exposed to prenatal SHS are at a higher risk of developing AD with its onset after age 2, and that screening for these high-risk groups may help prevent childhood AD earlier. Further studies are warranted to understand the underlying mechanisms.
A study in Japan reported no relationship between prenatal smoke exposure and the risk of AD in early childhood up to 3 years, which is consistent with our observations despite the study’s short follow-up period (16). Another prospective cohort study reported an association between prenatal smoke exposure and increased wheezing but decreased atopic eczema until age 3 (17). Our study investigated data from a longer follow-up period, allowing recognition of the late-onset manifestation of AD.
We found that the cumulative effect of SHS on AD was not apparent in early infancy and was only notable after reaching childhood. Additional analysis of AD phenotypes revealed that this effect is likely due to an increase in the late-onset AD phenotype, which develops after 2 years. While not statistically significant, an incremental relationship (p<0.1) was observed between higher cotinine levels and the late-onset AD phenotype. The prevalence of AD in Korea peaks during infancy and then decreases throughout early childhood (18), suggesting that AD aggravated by prenatal SHS may occur as the late-onset phenotype through a different mechanism from conventional AD.
Tobacco smoke induces the formation of hydrogen peroxide and activates the cellular NOX (nicotinamide adenine dinucleotide phosphatase oxidase), leading to the translocation and subsequent loss of SR-B1 or the HDL receptor. This may affect the stratum corneum, composed of 25% cholesterol (19). Tobacco smoke also exhibits oxidative effects in human skin fibroblasts (20). DNA methylation is reportedly induced by maternal smoking in pregnancy, which may mediate the effect of maternal smoking on AD (21). The methylation status of the TSLP 5’-CpG was significantly higher in the high-exposure group based on cord blood cotinine, and the degree of methylation was associated with decreased TSLP protein expression and increased AD (22). Hence, prenatal tobacco exposure may affect DNA methylation, leading to delayed AD occurrence. Only 0.23% of the mothers in the COCOA study reported smoking during pregnancy (data not shown). Therefore, a study focusing on the effect of SHS on AD will have high clinical significance in the Korean population.
The relationship between urine cotinine and AD was analyzed to determine the quantitative effect of prenatal SHS exposure on AD. The relationship between urine cotinine and smoking status (23) has been demonstrated, and a significant relationship between “smoking currently permitted in the whole house” and positive urine cotinine has been reported (9), indicating that maternal urine cotinine levels are a significant surrogate marker for SHS exposure. However, no significant relationship was observed between AD in early childhood (ages 0–3) and cotinine levels. The definition of AD in the earlier phase of childhood tends to vary, and a significant portion of patients undergo remission with various contributors. From this study, school-age (ages 7–9) data were insufficient for urine cotinine analysis, but a significant relationship was observed between AD in preschool children (ages 4–6) and urine cotinine levels during pregnancy, indicating an association between higher doses of cotinine and AD in childhood.
We applied the mediation model with SHS as exposure, offspring AD as the outcome, and IgE level as mediator (Supplementary Fig. 1). Total effect of SHS on atopic AD at school age (ages 7–9) was significant (OR = 2.033, p = 0.029). IgE level at age 3 significantly mediated the relationship (indirect effect OR = 1.110, p = 0.010, the proportion mediated = 14.8%), but the level at the other ages (age 1 or 7) had no indirect effect. These results showed that the IgE level at 3 years of age is a mediating factor in the relationship between SHS exposure and AD in sensitized school children. However, further study is warranted given that this association was not mediated by IgE level at other ages, and the indirect effect of IgE level was weaker than expected (Supplementary Fig.1.). Discussion regarding mechanisms related to IgE are in the online supplement.
There are a few limitations to this study. First, data on SHS exposure were investigated using questionnaires, and the intensiveness of the exposure was not considered. While it is typical to measure cotinine in the second or third trimester to assess the level of smoke exposure during pregnancy (24, 25), we measured urine cotinine at week 36 per the COCOA protocol. Nevertheless, exposure status to prenatal SHS is expected to be consistent through pregnancy since most exposure is expected to have occurred at home or work.
The main strength of our study is its prospective design. Data on the SHS exposure of pregnant mothers, their urine cotinine levels, and other potential confounders were investigated before birth, reducing biases that may corrupt data. An additional strength is that the assessment of AD was examined by pediatric allergists using a standardized research data form, and that phenotypes of AD were assessed. Furthermore, all children were adjusted for SHS exposure during their first year of life to distinguish the effects of prenatal and postnatal SHS exposure since the latter is also a major risk factor for AD(26). Children in the school age (7–9 years) group were adjusted for SHS exposure from ages 4 to 6. The COCOA cohort is a general population cohort, allowing generalization of the results of this study, especially in Asian countries with a low rate of maternal smoking during pregnancy.