Polychlorinated biphenyls (PCBs) have been widely used in industrial production, e.g. in capacitors, sealants, plasticizers, fire retardants and hydraulic fluids, but they were banned in the late 1970s in the United States and the early 1980s in Europe because of their environmental persistence [1,2]. Also, the insecticide dichlorodiphenyltrichloroethane (DDT) and its main metabolite dichlorodiphenyldichloroethylene (DDE) are highly persistent in the environment, and although the agricultural use of DDT has been severely restricted through the regional United Nations Economic Commission for Europe and the Stockholm Convention [3], it is still used as disease vector control in some developing countries [4]. PCBs and DDE are lipophilic compounds which bio-accumulate in adipose tissue and, thus, the main exposure is through food consumption and breast feeding [5,6]. The compounds pass the placental barrier and the fetus is exposed to approximately the same concentration of DDE as the mother when calculated on lipid basis [7,8]. The level of PCBs in the fetus, however, seems to be somewhat lower than that of the mother although correlated [9]. PCBs and DDE can be detected in serum and breast milk samples from almost all women around the globe, causing the fetus and subsequent breast-fed baby to be exposed to the compounds during infancy [9-12].

A major contamination of rice oil by heated PCBs and degradation products in Taiwan in 1979 gave insight into the negative impact of high prenatal exposure on e.g. neurodevelopment [13,14]. Animal studies later found decreased muscular strength [15] after high prenatal exposure to tetrachlorobiphenyl and reduced motor activity [16] after prenatal exposure to PCBs. However, results from epidemiological cohort studies are inconsistent: Higher in utero exposure to PCBs (3000- > 4000 μg/L) was related to a significant decrease in motor abilities indicated by 7–8 points decrease of the Psychomotor Development Index scores (PDI) in Bayley Scales of Infant Development (BSID) (a screening tool measuring children’s motor and mental abilities) compared with low exposure (0–900 μg/L) at six, 12 and 24 months in a birth cohort from North Carolina [17,18]. The Collaborative Perinatal Project from 12 hospitals in the United States found no association between background levels (0–16.50 μg/L) of prenatal PCBs exposure and motor development in terms of PDI in a sample of prenatally PCB exposed children of 8 months of age [19]. A two-point decrease in PDI scores for every 10-fold increase of DDE has been observed at six months of age, but not at 12 and 24 months [20], and further, no association was found between prenatal DDE and motor development according to McCarthy Scales of Children’s Abilities (MSCA) in two Spanish birth cohorts [21]. In addition, no association was found between in utero DDE exposure and motor development using BSID or MSCA among infants of one month, toddlers of 30 months or children of 3.5-5 years [22-24]. However, in utero exposure to DDE has been associated with hyporeflexia at one month in a cohort of North American children [25] and 4 points lower PDI for each doubling of DDE [26]. Further, no inverse associations between prenatal DDE and motor skills were observed in later follow-ups of the large North Carolina cohort [17,18].

In this cohort, we follow three populations of 1,103 children from Arctic and Europe with a large exposure span, which enables us to examine the relation between different levels of organochlorine exposures and child motor development in various parts of the world. Thus, the aim of this follow-up study was to examine the relations between prenatal exposure to p,p′-DDE and CB-153 and developmental milestones in infancy measured retrospectively and motor development assessed by the questionnaire “Developmental Coordination Disorder Questionnaire” in five to nine year old children in Greenland, Ukraine and Poland.

A total of 1,103 of the 1,441 baseline participants were included in the follow-up analyses (77%). Although the pregnant women had to be at least 18 years of age to be included in the study, for unknown reasons, three women were 16 years old and 11 women were 17 years old in Ukraine, calculated based on the recorded date for filling in the questionnaire and date of birth.

Crude results are presented as supplementary material (Additional file 2 and Additional file 3). Medium CB-153 levels was positively associated with age at standing and walking in the total population, corresponding to approximately two weeks delay, compared to low exposed children. No associations were observed in the high exposed group, in analyses stratified by country or between DDE and either outcome.

When restricting our analyses to subjects with complete information, we did not observe any associations between maternal serum CB-153 and p,p′-DDE levels and either of the child’s developmental milestones in any of the three countries or in the pooled analysis (Additional file 4) except for mean age at walking that appeared to be slightly lower at higher serum p,p′-DDE levels in Greenland (p for trend, 0.04). In Poland, medium CB-153 exposed children had a lower mean age at standing compared to low exposed children; −1.0 (CI: −1.9, 0.1) months. In the complete-case analysis, results of adjusted mean differences between tertiles of CB-153 and p,p′-DDE exposures and parentally assessed motor skills at young school age are similar to the results of the multiple imputation based analyses (Additional file 5).

Using breastfeeding duration as the exposure in a sub-analysis, we found no relation to motor scores (data not shown). Excluding 37 preterm children in a sub-analysis gave roughly the same results except for a borderline statistically significant difference in Greenland. High DDE exposed children walked 0.7 months later than low exposed children (−0.7 (95% CI:-1.4; 0.0)).

The four different sensitivity analyses performed on the imputation model did not change the direction or the magnitude of the estimates (data not shown). Also, stratification by maternal education gave no clear dose–response association according to maternal education. When testing for interaction between the exposures and country, no apparent signs of interaction were found.

We investigated whether in utero exposure to p,p′-DDE and CB-153 was related to parentally retrospectively assessed delayed developmental milestones as age at crawling, standing-up and walking in infancy, and motor skills at 5–9 years of age as measured by DCDQ. We did not observe any adverse association of prenatal exposure to CB-153 and p,p′-DDE on developmental milestones or motor development.

Our results are consistent with some earlier studies with generally lower exposure levels [19,23,33]. Daniels et al. reported no association between PCBs (median (95% CI): 2.7 (1.8, 3.7) ng/ml) and motor skills in terms of BSID among 8 months old infants [19], in line with a Canadian study among 5 year-old Inuits (CB-153 mean: 99.6 ng/g lipid) [33] and a Mexican study of DDE (mean: 6.3 ng/ml) and motor skills in 30 months old children [23].

Studies with large exposure contrast between studies have found negative associations between in utero exposure to either or both of the compounds and parts of the motor development [17,20,25,26,34,35]. Some of the early studies on very high levels of maternal DDE (range: 0- > 6 mg/L in milk fat) and PCBs (range: 0- > 4 mg/L in milk fat) exposure and motor development in the offspring observed associations between PCBs and statistically significantly lower motor developmental scores according to BSID and between DDE and hyporeflexia according to Brazelton Neonatal Behavioral Assessment Scale in the same cohort [18,25]. The discrepancies in results could be due to the higher exposure levels and their use of other, and more detailed, outcome measure.

Among 92 toddlers, prenatal exposure to DDE has also been associated with lower PDI scores in Spain [26]. The age difference between the children of the cohorts and our use of a less detailed outcome measures could be the reason for the disparate results. A study found a 10-fold higher prenatal DDT exposure to be associated with a two-point lower psycho-motor development BSID score at six and 12 month old infants. High DDE-levels (geometric mean 1,437 ng/g lipid) were also associated to lower psycho-motor development scores at six months but not at 12 or 24 months, indicating DDT to be a psycho-motor toxic compound as well as DDE [20]. Mean DDE-levels in the study by Eskenazi et al. is more than twice the level of the highest mean level in our study and their use the more detailed BSID questionnaire, may explain the dissimilarities in results.

Breastfeeding may potentially confound the results due to its association to postnatal exposure to PCBs and DDE. On the other hand, recent studies have found breastfeeding beneficial in relation to motor development, seemingly counterbalancing the possible negative consequences of DDE [20,26]. In our data, breastfeeding does not seem to be a risk or protective factor for motor development neither in general nor when stratified by prenatal exposure levels (data not shown). We did not intend to evaluate the associations between postnatal CB-153 or DDE exposure and motor development in the present study.

In this study, the blood sample was drawn in the second or third trimester of pregnancy as did many others [19,20,34,36]. As the half-life of both compounds of interest is several years in the human body it seems unlikely that large differences in concentration appear between trimesters of pregnancy. However, high levels of first trimester DDE has been associated with motor development in infants till the age of 12 months, whereas no association was seen between second and third trimester DDE and motor development [37]. This may be a chance finding since their study, to the best of our knowledge, is the only one reporting such results. Also, Longnecker et al. found the level of DDE to be highly correlated through the trimesters of pregnancy and they concluded that the timing of measurement was not important [38].

The three diverse populations of this study have enabled us to follow children in different settings with a large exposure span. The long follow-up of up till 9 years is important since it is of importance to disclose whether possible associations between prenatal organochlorine exposure and motor skills persist into school age. In addition, these associations have rarely been investigated in these populations.

This study has some limitations. Developmental milestones were assessed retrospectively, introducing the possibility of information bias. However, we have no reason to believe that differential misclassification by exposure levels has affected the estimates as parents were unaware of their exposure levels when answering the questionnaire. However, non-differential misclassification may have attenuated the true association.

We did have a considerable proportion of missing data in the dataset, in particular on the total DCDQ-score in Greenland. In the Greenlandic version of the questionnaire (as opposed to the Danish version) one item lacked due to an error, making a calculation of the total score impossible in the Greenlandic version before imputation. Thus, we performed chained multiple imputation on the dataset because of concern for selection bias caused by missing information. Analyses on a full dataset with no missing items would have been preferable, however, when missing data does occur and data is missing at random, multiple imputation is superior to complete case analyses [39].

The median DCDQ-score was lower in Ukraine than in Greenland and Poland. This may be due to cultural differences among study sites, which has been previously reported [31]. The reliability and validity of DCDQ is acceptable [40] although previous studies have indicated that the original DCDQ may have poor sensitivity in a population sample compared to the physical motor test The Movement Assessment Battery for Children (MABC) [41] and also, the later version, DCDQ’07, had low specificity when screening an adolescent population [42]. As we intended to examine whether there was a difference in motor score according to exposure levels of DDE and CB-153, we do not believe this is of major concern for our study. In spite of a reported agreement of between 80 and 95% between DCDQ and expert opinion [40] interpretation of the prevalence of DCD in the various countries is not recommended without an additional standardized motor test to confirm the screening result of DCDQ.

Because of a relatively small study population in Poland and a substantial amount of covariates in the regression model, the results from Poland should be interpreted with caution. In Poland, the participation rates at baseline and at follow-up were rather modest (37% and 36%, respectively) and thus, the Polish part of the cohort may not fully reflect the reference population. In Ukraine, the baseline participation rate was quite low (25%) but very high at follow-up (81%). The participation rates in Greenland were very high both at baseline (88%) and at follow-up (89%) suggesting a good representation of the reference population in Greenland. Unfortunately, we were not able to compare non-participating and participating women at baseline. Comparison of respondents and non-respondents at follow-up indicate no large differences, however, loss to follow-up in particular in Poland, may have caused selection bias.

The relatively high correlation between CB-153 and DDE made us choose not to perform mutual adjustment in the analyses, and thus possible confounding of the opposite organochlorines in the results could not be eliminated.

Smoking and drinking alcohol prior to pregnancy were used as proxies for the habits during pregnancy as this information was unavailable. This may have lead to confounding of the estimates, as the women may have changed their behavior during pregnancy.

Omega-3 fatty acids may be positively related to motor development [43] and would co-occur with the exposures. Unfortunately we have no specific information on omega-3 fatty acid consumption, and thus, we cannot rule out the possibility of omega-3 fatty acid counterbalancing the possible negative effects of the organochlorines.

The results of this follow-up study show consistently no indication of an adverse effect of prenatal exposure to CB-153 and p,p′-DDE on motor development in Arctic as well as in European countries representing countries with markedly different exposure levels. However, the lack of a relation between prenatal exposures to CB-153 and DDE and motor skills may be due to generally low exposure levels or a lack of variability in range of the exposures or both, compared to earlier studies. Also, developmental retrospectively recalled milestone may not be sensible enough to detect a difference according to exposures at these levels. The use of current motor skills (DCDQ) at school age, on the other hand are not prone to recall bias.

Prenatal exposure to CB-153 and p,p′-DDE were not associated with the parentally retrospectively assessed developmental milestones, age at crawling, standing and walking in infancy or parentally assessed motor skills in young school age children from Kharkiv (Ukraine), Warsaw (Poland) and Greenland, indicating no major effect of CB-153 or p,p′-DDE exposure on these outcomes. However, subtle effects cannot be excluded and the use of a more sensitive outcome measure may be warranted.