Between 1996 and 1999, 325 primiparous women, living and seeking prenatal care in Uppsala County, Sweden, agreed to donate a serum sample (82% participation rate) in late pregnancy (week 32–34) for chemical analysis of organochlorines 19. PCB and DDE concentrations in serum lipids from this sampling were used as an estimate of prenatal PCB and DDE exposure of the infants, and 190 mother/child pairs had complete data for statistical analyses of associations between prenatal PCB and DDE exposure and diseases among the infants (Table 1).

During the third week after delivery mothers from the group of 190 participants sampled milk while nursing their infants, using a manual breast pump and/or a passive mother's milk sampler 20. The results of the PCB and DDE analyses of mother's milk was used in the assessment of postnatal exposure (see below) and 175 mother/infant pairs had complete data for statistical analyses of associations between postnatal PCB and DDE exposure and risk of infections.

Blood was sampled from infants 3 months after delivery for analysis of numbers and percentages of WBCs and lymphocyte subsets. The age of 3 months was chosen in order to facilitate comparison with other studies of immunomodulatory effects of early life PCB and DDE exposure 1113. Moreover, the majority of the infants in Sweden have mothers'milk as the only source of nutrition during the first 3 months of life, making confounding of results due to differences in sources of nutrition less pronounced. All the 325 mothers, donating blood in late pregnancy, were asked if they were willing to donate an infant blood sample after birth. Blood sampling could however not be continued during the whole study due to lack of financial resources. Therefore infants of mothers accepting blood sampling were sampled until 90 infants had been sampled. At that time point 277 mothers had been asked to donate a blood sample from their infants and the participation rate in the blood sampling of the infants was thus 32%. Among the 90 mother/infants pairs, data on numbers of WBCs were finally available from 81 infants and percentage data from 85 infants. Lymphocyte subsets could only be analyzed in 52 infants from this subgroup of infants, due to limited volumes of blood available. Among the infants with data on lymphocyte subsets, data on lymphocyte numbers and percentages were available for 47 and 52 infants, respectively. The study was approved by the Ethics Committee of the Medical Faculty at Uppsala University. All participating women gave their informed consent prior to the inclusion in the study group.

At 6–12 and 32–34 completed gestational weeks, in-person interviews regarding maternal characteristics were performed, using a structured questionnaire 1921. Data on maternal characteristics included age, height, body weight before pregnancy, body weight at interview, years of education, and alcohol consumption and smoking before and during pregnancy. Blood samples for cotinine analysis (used as indicator of smoking habits) were taken at both interview occasions.

After delivery, the mothers were visited by a midwife when the infant was 3 months old. In an in-person interview, using a structured questionnaire, the mothers answered questions about sex of the infant, vaccination, nursing habits, and infant health during the first 3 months. Women gave information about the extent of nursing for each week of the 3 month period (full nursing, partial nursing and no nursing). The participants also gave information about the health of the children each week from delivery to the date of the interview. It was asked if the infant had any infection or other disease during the 3 month period. If the answer was yes, the mother was asked to give information about type of disease (open ended question). The mother was asked to identify the week/weeks the infant had the disease in an almanac, and to give the number of days the symptoms lasted. It was also asked if the infant had a fever during these days, and number of days with a temperature above 39°C. Finally it was asked if the symptoms were treated by a physician and what type of treatment that was used. The interviewer and the mothers were blinded to the organochlorine compound body burdens of the mother.

The lipid portion of serum samples was analyzed for the DDT metabolite p,p'-DDE, and 10 PCB congeners (IUPAC nos. 28, 52, 101, 105, 118, 138, 153, 156, 167 and 180). The analytical results were consequently lipid adjusted. Procedures for extraction, sample clean-up and analysis, and quality control are described in Glynn et al. 19. In mother's milk the compounds were analyzed using a method and quality assurance described by Glynn et al. 20. When concentrations were below the limit of quantification (LOQ) they were set to 50% of LOQ in the statistical analysis.

The hematological tests were performed at the Department of Clinical Chemistry, University Hospital of Uppsala. Capillary blood samples were collected in microtainer tubes (Sarstedt, Sweden). Differential counts were carried out by an automated instrument, Celldyn 4000 (Abbott Scandinavia AB, Solna, Sweden) based on a combination of optical characteristics and histochemical reactions. Number of total WBCs, and number and percentages of neutrophils, eosinophils, lymphocytes, and monocytes were recorded.

Analysis of lymphocyte subpopulations was performed on mononuclear cells prepared by Ficoll-Paque centrifugation and stained as described in Gräske et al. 22. The following subpopulations were evaluated by flow cytometry: CD3⁺ cells (T-lymphocytes), CD19⁺ cells (B-lymphocytes), CD4⁺ cells (T-helper cells), CD8⁺ cells (cytotoxic cells), and CD56⁺ cells (NK-cells).

Lipid-adjusted serum organochlorine compound concentrations among the mothers in late pregnancy were used as a measure of prenatal exposure of the fetus 232425. Postnatal exposure (PE) (ng or pg/g*days) was calculated from the organochlorine concentration (ng or pg/g fresh weight) in mother's milk three weeks after delivery (Oconc), the number of days of nursing (Nd), and the percentage of full nursing during the study period (%N) using the equation

PE (ng or pg/g*days) = Oconc*Nd*(%N/100).

For women who did not nurse their infants PE = 0. The nursed infant's cumulative intake of organochlorine compounds is the product of the concentration of the organochlorine compound in question in the mothers' milk (per gram fresh weight), the amount of milk consumed per day (in grams), and the number of days of nursing 2627. We did not have information about the amount of milk consumed by the infants. Instead we had information about the degree of nursing each week (full, partial or none). Consequently, the degree of nursing was used instead of the amount of mothers' milk consumed in the intake calculation.

In the exposure analysis, tri- to penta-chlorinated CB-28, CB-52 and CB-101, with no dioxin-like biological activity, were grouped together (CB 28+52+101) because serum levels of these compounds showed low correlations with serum levels of the other organochlorine compounds (Spearman's r = 0.13–0.24). These three congeners have been detected in elevated levels in indoor air of buildings containing PCB-laden building materials and in the blood of residents of such buildings 28293031. CB-138, CB-153 and CB-180 (di-ortho PCBs) were grouped together because of high correlations between mother's serum concentrations of these congeners (r > 0.90), and because of their lack of dioxin-like biological activity 32. Dioxin-like mono-ortho PCBs consisted of CB-105, CB-118, CB-156 and CB-167 32. p,p'-DDE was treated separately in the statistical analysis. In order to improve the possibilities to compare our PCB results with those of other similar studies, statistical analysis was also performed using exposure to the di-ortho PCB congener CB-153 as independent exposure variable. Concentrations of CB-153 in blood serum of the mothers was highly correlated with concentrations of total PCB (Spearman's r = 0.97), showing that CB-153 exposure is a good marker of total PCB exposure.

Exposure to dioxin-like mono-ortho PCBs were summarized using WHO 1998 toxic equivalency factors (TEFs) for the congeners CB-105 (TEF = 0.0001), CB-118 (0.0001), CB-156 (0.0005), and CB-167 (0.00001) 32. The concentration of the individual congeners was multiplied with the TEF of the congener in question 32, and the resulting concentration of toxicity equivalents (TEQs) for each congener was then summarized in to a sum of mono-ortho TEQ concentration for each sample.

Statistical analysis was performed using MINITAB^® For Windows, 14. Spearman's rank correlation analysis was used in analysis of correlations between different exposure variables. The associations between immune cell numbers/percentages and organochlorine compound exposure were explored by linear regression analysis. Six infants had an ongoing infection at the time of sampling, or an infection within a period of 7 days before the sampling day. These infants were not included in the statistical analyses. Regression analysis was performed on logarithmically transformed organochlorine compound exposure data, since the distributions of data closely followed a log-normal distribution. However, in the case of prenatal exposure to CB 28+52+101 many women had serum levels below LOQ and this exposure variable was therefore categorized. Some women did not nurse their infants at all, and postnatal exposure of these infants was set to zero. This made logarithmic transformation impossible and postnatal exposure was therefore categorized. Prenatal CB 28+52+101 exposure was categorized in 3 categories. The study participants could not be categorized in tertiles since over 40% had concentrations below the LOQ. Mother/infants pairs with CB 28+52+101 concentrations below the LOQ were grouped in the reference category and the rest of the study participants divided up in equal numbers in the two other categories depending on exposure levels. Postnatal organochlorine compound exposure was categorized in tertiles when possible, otherwise an effort was made to have equal numbers of study participants in each of the three exposure categories.

In the statistical analysis of immune cell results the significance level was set to p ≤ 0.01, since multiple comparisons were made. In cases when a statistically significant association between immune cells and organochlorine compound exposure was found in simple regression analysis, multiple regression analysis was used to adjust the associations for potential confounders. We included lifestyle/medical factors in the multiple regression analyses that have previously been reported to be associated with increased or decreased risk of immune-related diseases 333435363738. Potential confounders included in the analysis were age of the mother, smoking during pregnancy (non-smoker/former smoker/smoked during pregnancy), alcohol consumption during pregnancy (no/yes), mother's education (≤ 13 years of education/14–16 years of eduction/>16 years of eduction), vaccination status of the infant at the time of sampling (no/yes), nursing of the infant (no or partial nursing/full nursing), age of the infant, and infant's past history of respiratory infections during the study period (up to 7 days before sampling, no/yes). In linear regression analysis a few observations with a standardized residual ≥ 3.0 were omitted from the data sets due to a large impact on the regression results.

In the analysis of infection results (no/yes), odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using logistic regression. Respiratory infection (influenza-like symptoms, common cold, or common cold with cough) (N = 54) was the only health problem that was frequent enough to allow statistical analysis. Other health problems reported were stomach problems (N = 1), navel infection (N = 1), cows' milk intolerance/allergy (N = 1), urinary infection (N = 1), eye infection (N = 2) and middle ear infection (N = 2). In the logistic regression analysis all independent variables were categorized in order to handle outlier problems. First, ORs were calculated in regressions with the exposure to each organochlorine compound group as the only independent variable. In the next step the ORs were adjusted for the same possible confounders as in the analyses of immune cell counts/percentages (except infant's history of infection, see above).

A stricter definition of respiratory infection was also used (three days or more of infection, N = 50) in the statistical analysis. The information about number of infections during the study period could not be used in the statistical analyses of the results, since only 7 infants had more than one infection during the study period. Similarly, too few infants had been examined by a physician. It was not possible to use the information about infection in combination with information about body temperature in the statistical analysis, since there were too many missing values for the latter variable.

We also analyzed the associations between respiratory infections and the total exposure to PCBs and p,p'-DDE during the whole prenatal and postnatal period. This analysis was based on a summation of each infant's categorized pre-and postnatal exposure to all compound groups. We did not calculate a sum of the absolute concentrations since such a summation assumes equal immunomodulatory potency of all the studied compounds. For example, when absolute concentrations are summed up, effects of relatively potent compounds present at low concentrations may be underestimated, and effects of less potent compounds present at high concentrations overestimated. We avoided this by summing up exposure scores for each compound group. First each infant's pre- and postnatal PCB and p,p'-DDE exposures were separately scored, with the lowest score 1 given to the lowest exposure quartile and the highest score 4 given to the highest exposure quartile (scores 1–3 for prenatal CB 28+52+101 exposure). The scores for pre- and postnatal exposures were then summarized for each compound group. For example, if an infant scored 1 for prenatal exposure to p,p'-DDE and scored 3 for postnatal p,p'-DDE exposure then the infant was assigned a score of 4 in the summation of pre- and postnatal p,p'-DDE exposure. The resulting sum of the pre-and postnatal exposure scores of each compound group was then added to get the total exposure score. In this case, if an infant had scored 5 for total CB 28+52+101 exposure, 8 for mono-ortho PCB exposure, 6 for di-ortho PCB exposure, and 5 for p,p'-DDE exposure, the total organochlorine compound exposure score was summed up to 24. The total exposure scores were subsequently divided into quartiles in the final statistical analyses of associations between infections and total organochlorine exposure. The level of statistical significance in logistic regression analysis was set to p ≤ 0.05.

Table 1 gives the personal characteristics of the whole study group (infection), the subgroup of mother/infant pairs that donated blood for WBC counts, and mother/infant pairs within this subgroup that donated enough blood for lymphocyte subset analyses. A comparison of personal characteristics between the study participants that did not donate infant blood (N = 104) with those donating infant blood (N = 86) showed no significant differences in age of the mothers and infants (t-test, p = 0.29 and p = 0.82) (Table 1). No difference was found between the two subgroups in smoking during pregnancy (Chi-square test, p = 0.064), alcohol consumption during pregnancy (Chi-square test, p = 0.89), years of education (Chi-square test, p = 0.57), nursing (Chi-square test, p = 0.28), and respiratory infections (Chi-square test, p = 0.52). Furthermore, pre- and postnatal organochlorine compound exposure did not differ between the groups (t-test, p = 0.054–0.84). However, there were more girls than boys in the subgroup with blood samples, whereas the reverse was evident among mother/infant pairs that did not donate infant blood (Chi-square test, p = 0.002). Fewer had been vaccinated at the end of the study period among infants with blood samples (Chi-square test, p = 0.038).

The median prenatal exposure of the sum of CB-28, CB-52 and CB-101 was low (Table 2). Over 40% of the mothers in the whole study group had concentrations of all the three substances in serum lipids below the LOQ. Concentrations ≥ 10 ng/g lipid were however found in 15% of the study participants. Median concentrations of di-ortho PCBs (CB-138, CB-153 and CB-180) and p,p'-DDE were similar, and the concentrations varied 8- to 26-fold. The concentrations of mono-ortho PCB TEQs varied 9-fold or more (Table 2).

Correlations between serum concentrations of CB 28+52+101 and the other organochlorines in the total study group (infection) were weak, with Spearman correlation coefficients ranging from 0.13 to 0.24. Concentrations of CB-153, di-ortho PCBs and mono-ortho PCB TEQs were strongly correlated (r ≥ 0.93), whereas correlations between these PCBs and p,p'-DDE were less strong (r = 0.66–0.75).

Variation in organochlorine exposure from mother's milk during the first 3 months of infancy was large (Table 2). The correlation between postnatal exposure to CB-153, di-ortho PCBs and mono-ortho TEQs was strong (r ≥ 0.93). Other correlations between the studied compound groups were less strong (r = 0.40–0.82). Pre- and postnatal exposure of each compound group was significantly correlated, ranging from r = 0.71 to r = 0.78.

The numbers and percentages of WBCs are presented in Table 3. In both simple and multivariate regression analysis, infants with the highest CB 28+52+101 exposure had significantly higher mean numbers of total WBCs, lymphocytes and monocytes than infants in the reference category with the lowest exposure (Figure 1). The percentage of eosinophils was negatively associated with prenatal p,p'-DDE exposure (Table 4). Otherwise, no significant associations between prenatal exposure and WBC counts/percentages were found (Figure 1, Table 4).

Numbers and percentages of WBCs were not significantly associated with postnatal exposure to any of the analyzed organochlorines, except for a higher percentage of lymphocytes among infant with the highest post-natal p,p'-DDE exposure in the univariate analysis (see Additional file 1 and Additional file 2).

We found a statistically significant negative association between percentage of CD4^-CD8⁺ cells and prenatal exposure to CB-153, di-ortho and mono-ortho PCBs in the univariate analysis (Table 4). After adjustment for potential confounders the associations were not statistically significant. In other cases, numbers and percentages of different lymphocyte subsets were not associated with prenatal exposure (Figure 2, Table 4). Postnatal exposure of the infants to the organochlorines was not significantly associated with numbers and percentages of lymphocyte subsets, except for a significantly higher number and percentage of CD56⁺ among infants in the second highest exposure group of post-natal CB-153 exposure (see Additional file 3 and Additional file 4).

Infants with the highest prenatal CB 28+52+101 exposure had a significantly increased odds ratio for respiratory infections during the first three months after birth compared with the lowest exposed infants (Table 5). This finding did not change in multivariate-adjusted logistic regression analysis. Moreover, a lowered odds ratio was found for infants in the second to fourth exposure category of prenatal di-ortho PCB exposure, although the difference between the highest exposure category and the reference category only showed borderline statistical significance (Table 5). Similar results were found for prenatal CB-153 exposure. A lowered odds ratio was also found for infants in the highest exposure category of prenatal mono-ortho PCB exposure in the multivariate-adjusted analysis (Table 5). Infants with the lowest prenatal p,p'-DDE exposure had a higher odds ratio than infants with higher p,p'-DDE exposures, but the difference never reached statistical significance. Similar results were evident when a stricter definition of respiratory infection (3 days or more of infection) was used (Additional file 5).

In the analysis of postnatal exposure, the OR for respiratory infections among infants in the highest exposure category of CB-153 and di-ortho PCB exposure was significantly lower than that of the reference category in the unadjusted analysis (Table 5). Multivariate analysis did not change this difference markedly. Significantly lowered ORs were observed for the second exposure category of p,p'-DDE and mono-ortho PCB, suggesting a U-shaped relationship (Table 5). An U-shaped relationship was also suggested for CB 28+52+101, but the differences in odds ratio between the reference category and the other exposure categories were not statistically significant. Similar results were found when a stricter definition of respiratory infection was used (infection for 3 or more days) was used in the statistical analyses (Additional file 5).

ORs for respiratory infections among infants in the second, third and fourth categories of total organochlorine compound exposure during the whole pre- and postnatal period were 0.74 (95% confidence interval:0.29–1.88), 0.36 (0.13–0.96), and 0.68 (0.28–1.68), respectively.