In this analysis, we saw an association between HCB and birth weight, both when examined continuously and when using indicators of fetal growth restriction. Our analyses suggested that this association may, however, be restricted to current and past smokers. We saw a similar pattern for the other anthropometric measures. A stronger effect of environmental toxicants on birth weight among smokers has been reported for PCBs (14
), but this is the first study to note an effect modification by smoking for HCB. Smoking is a well known risk factor for low birth weight and, as such, may be important to consider as an effect modifier. Cigarette smoke contains a number of toxins, such as cadmium, cyanide, sulphides, carcinogenic hydrocarbons, benzene, and nicotine, which all may cause direct cellular damage as well as specific biological effects, as recently reported (29
). Still, the specific causal mechanism for the association between smoking and birth weight remains unknown. A mechanism for the modifying effects of smoking, if not a chance finding, is therefore unclear.
The association between HCB and the reproductive outcomes examined was unexpected, considering the relatively low levels of HCB recently observed in human milk in Norway. This point to the possibility of a non-causal association. An underlying genetic variant or an unknown environmental factor that is involved in both the breakdown of HCB and fetal growth could account for the observed inverse association (30
). However, it is also possible that the relevant exposure window for HCB on these outcomes may have been at an earlier time point, such as during puberty. The women in this study would have been exposed to much higher HCB levels in their youth, because levels were much higher in the past (3
). The current ranking of women for HCB levels may simply reflect how they would have been ranked in the past. A recent study concluded that “a single organochlorine measure provides considerable information on relative ranking at distant times” (33
), supporting the possibility that the relative ranking of HCB may be stable over time. This hypothetical explanation could, if true, account for the variation in results across studies, since past HCB exposure levels may differ between countries even when recent levels were similar.
Another potential explanation is the possibility of differential effects on outcome depending on exposure dose. Protective mechanisms could be activated at higher levels, preventing the negative effects that are seen at lower levels (34
). However, the GAM model did not reveal any significant threshold effects on birth weight across the range of HCB seen in this study. This is in accordance with the only other study that has looked for a non linear relationship, which was also performed in a population with low HCB levels (14
). Although there may be threshold effects at higher HCB levels, there are no studies to date to either support or refute this possibility.
The results for gestational age are difficult to interpret. We saw a suggestion of a nonlinear relationship between categorical HCB and gestational age. However, when the one subject with high influence was removed, the relation appeared to be linear. Overall, we cannot draw conclusions on the relation between HCB and gestational age in these data, in part due to the imprecision of the estimates.
Overview of previous findings
The relation between HCB levels and anthropometric measures at birth has been examined in 7 previous studies (3
). The highest exposure levels were reported in a study from Spain including 70 subjects, with HCB measured in cord blood (10
). No association with birth weight or SGA was observed. However, the authors did note a statistically significant inverse association with crown- rump length, suggesting impaired growth (10
). Among 107 Inuit children with relatively high HCB levels measured in human milk, an association between HCB and reduced length at birth was also reported, but no association was seen for birth weight (35
). In a third study of 197 subjects, still with relatively high levels of HCB measured in milk, no association with birth weight was observed (12
). Length was not assessed in this study.
There are two studies with intermediate levels of HCB. HCB measured in human milk in 246 subjects was associated with lower birth weight in a study from Germany (3
). In a study of 385 Latino-workers in the US (11
), in whom HCB was measured in maternal serum at approximately 26 weeks of gestation, a decrease in birth weight of 96 g per log10
-transformed units HCB μg/kg lipid was observed in the crude analysis, an estimate comparable to the crude associations in our study of 109 g per 8 units of HCB. However, when adjusting for maternal weight gain during pregnancy and other factors, the association was reduced to −23 g and was no longer significant. This study observed a significant association between HCB and preterm birth.
In a cross-sectional study of 815 Australian mothers, with levels slightly higher than ours, the associations with the outcomes were explored after categorization of HCB exposure into tertiles (15
). A twofold increased risk of low birth weight (OR 2.6, CI 0.96–7.3) was observed for those in the middle category compared to the lowest exposure category, the two categories which corresponded to the range of HCB in our study. No association between the highest exposure category and low birth weight were seen, which could suggest a threshold effect. However, no association with SGA was noted for any category of HCB exposure. Finally, in a study of 722 healthy newborns from the US, where HCB was measured in cord blood (14
), levels were comparable to our study, although they had a high proportion of non-detects (detection level equal to median level). No association between HCB and the anthropometric outcomes studied was present in this study.
Limitations and Strengths
The main limitation of our study is the retrospective assessment of exposure, although this is generally not raised as a major concern in studies on chlorinated persistent organic pollutants, due to their long half lives (36
). Postnatal HCB exposure is assumed to represent prenatal exposure, and for most persistent organic pollutants the levels in human milk and cord blood do show a very high correlation. However, our knowledge with regard to HCB specifically is more limited (37
Maternal recall of pregnancy-related events have been shown to have a high degree of validity (38
). We evaluated potential recall bias for smoking, the covariate probably most likely to be affected by recall bias, by reanalyzing all models using prospective smoking data and we observed very similar results.
A weakness of our study is the low final response rate. Women with higher education were more likely to participate in the study, which probably also explains the lower proportion of smokers, since smoking is associated with lower socioeconomic status in Norway (). The results may therefore not be generalizable to the whole population. However, we believe that it is not likely that the validity of the results is affected by selection bias, since that would require differential selection that is related to both HCB and to SGA,. SGA status could affect selection to the study, since SGA babies may need a prolonged stay at an intensive care unit and thus miss the visit by the health visitor by whom recruitment took place (although the birth weight is not very different in the study sample compared to the whole population ). However, we find it unlikely that mothers HCB levels, which are unknown to them, would directly be associated with selection to the study.
This study has some important advantages compared with studies that have used serum or cord blood to assess exposure. First, human milk enables sampling of large volumes. We used 15 ml for the contaminant analysis, an amount usually not available from either cord blood or maternal serum. The large volume and the high fat content of human milk allow measurement with greater precision, especially for contaminants that are present in low concentrations. This is important, since the varying fat concentrations in biological samples of either type imply that a sample with a non-detectable value cannot be assumed to have a lower concentration than a sample with a detectable value. Furthermore, to the best of our knowledge, this is the first study in which milk has been sampled over a prolonged period of time (eight days). This was done in order to reduce the high within-subject variability that has been demonstrated for these contaminants in lactating women (39
Another advantage of the present study was the availability of data on coexisting environmental contaminants. In all the adjusted models, the estimates were higher after inclusion of selected environmental contaminants, and the stepwise approach revealed that this effect was primarily tied to including PCBs into the model. Since environmental contaminants coexist in exposure sources and may have additive or antagonistic effects, this finding points to the importance of studying a broad array of contaminants that coexist in a specific population. Most studies on HCB have not adjusted for PCBs. Still, adjustment for other contaminants can only partly be achieved, and we cannot exclude the possibility of confounding by other as yet unmeasured persistent organic pollutants that are strongly correlated with HCB exposure in Norway.
In conclusion, we observed an association between HCB levels measured in maternal milk and size at birth, which was, however, restricted to current and past smokers. This association may be non-causal, or may reflect past exposure. Compromised fetal growth has been associated with increased risk of a number of diseases later in life, and understanding what mechanisms lay behind this association has important public health implications (40
). The observed effect modification by smoking is intriguing and, if confirmed, may increase our understanding of the mechanisms of action of these contaminants.