Study population
This study utilized the Swedish Environmental Longitudinal, Mother and child, Asthma and allergy (SELMA) study, a pregnancy cohort with the primary aim to investigate the importance of early life exposure to environmental toxicants with focus on endocrine disrupting chemicals (EDCs) during the pregnancy and infancy period for health and development in the children [25]. Pregnant women were recruited from September 2007 to March 2010 during their first visit (median = 10 weeks age of gestation) at antenatal care centers in Värmland, Sweden. Out of 8394 women, 7119 were invited and 2582 (39%) consented to participate. Maternal blood was collected, and questionnaires were used to gather information regarding their medical history, lifestyles, and socioeconomic status. Questions on TTP were asked by midwife at enrollment. From the original cohort, only women with complete data on TTP, serum POP concentrations, age, parity, preventive method, regularity of menses, BMI, and lifestyle factors were included in this study (n = 818) (Supp Fig. 1).
Time-to-pregnancy (TTP)
Fecundability or the probability of pregnancy in each cycle was measured through TTP, a sensitive and convenient marker used to study environmental exposures. Women were asked “How long have you been trying to get pregnant?” with answers in years and months. Additional questions on pre-pregnancy contraceptive use and reproductive history were also inquired. They were asked “What was the preventive method you used before you became pregnant?” and when they stopped using it. They were also asked if they have regular or irregular menses.
Chemical analyses of POPs in serum
Concentrations of 22 POPs (nine OCPs namely pentachlorobenzene (PeCB), hexachlorobenzene (HCB), three isomers of hexachlorocyclohexane (α–HCH, β–HCH, γ–HCH), oxychlordane, transnonachlor, dichlorodiphenyltrichloroethane (p,p´-DDT), dichlorodiphenyldichloroethylene (p,p´-DDE), ten PCB congeners namely, PCB 74, 99, 118, 138, 153, 156, 170, 180, 183, and 187 as well as three PBDEs namely PBDE 47, 99, and 153) were analyzed in the blood samples collected during the first healthcare visit of current pregnancy, median 10 weeks age of gestation, as described previously [26]. Briefly, 200 ul of serum sample were enriched with 400 pg 13C-labelled internal standards of each compound. POPs were extracted with 2 ml dichloromethane-hexane (1:4). Extracts were cleaned with multilayer silica columns and the eluate was concentrated for gas chromatography – tandem mass spectrometry (GC-MS/MS) analysis (Agilent 7010 GC-MS/MS system, Wilmington, DE, USA). Control serum sample (NIST SRM 1958) and an in-house low-concentration control sample (1 to 9 dilution of SRM 1958 with new born calf serum) were included. Concentrations were reported as wet weight (pg/ml) and the limits of quantification (LOQ) ranged from 5 to 40 pg/ml. LOQ was defined as the concentration corresponding to ten times the standard deviation of the signal-to-noise ratio. Of the 22 POPs analyzed, 13 were detected in more than 70% of the included women.
Analysis of cotinine in serum samples
Cotinine was measured in serum samples and used as a biomarker for tobacco smoke exposure. Isotopically labelled internal standards were added to 100 μl serum. Samples were digested with glucuronidase and proteins were precipitated using acetonitrile. Analyses were performed using triple quadrupole mass spectrometry (QTRAP 5500; AB Sciex, Foster City, USA) coupled to a liquid chromatography system (UFLCXR, Shimadzu Corporation, Kyoto, Japan) (LC-MS/MS). Subjects were categorized as non-smokers for cotinine concentrations below 0.2 ng/ml, passive smokers with cotinine concentrations 0.2–15 ng/ml and active smokers for cotinine above 15 ng/ml [27].
Statistical analyses
POPs detected in less than 70% of the samples were not included in the analyses. Among those chemicals detected in more than 70% of the mothers, concentrations below the LOQ were replaced with LOQ/2. Characteristics of the cohort were presented as n (%), unless otherwise specified. Chi-square test, Wilcoxon rank-sum test and t-test test were used to analyze demographics and chemicals among groups. Correlation among chemicals were determined using Spearman’s rank correlation. To compare exposure patterns between Sweden and the United States, the geometric mean of serum concentrations of the 13 POPs were used to determine cumulative burden and its relative proportion in the SELMA cohort as well as in women aged 20–39 years in the National Health and Nutrition Examination Survey (NHANES) in 2007–2010 conducted by Centers for Disease Control and Prevention in the United States [28, 29].
Analyses were done on the association of POPs on the outcome variable TTP, both as continuous (measured in months) and binary (infertile for those with TTP > 12 months). Chemicals were analyzed both as continuous (log10 transformed) and quartiles. Discrete-time Cox regression models were used to estimate hazard ratios of fecundability (FR). The proportional hazards assumption was met based on the statistical tests on the scale Schoenfeld residuals. The FR represents the per-cycle probability of conception in subgroups with higher concentration POP exposure relative to a reference group with lower concentration POP exposure [30]. We censored the data when TTP exceeded 12 months. Sensitivity analyses were done with censored data at 10 and 14 months. An FR > 1 denotes higher fecundability, which means shorter TTP and FR < 1 denotes lower fecundability, which means longer TTP. Moreover, logistic regression was used to estimate odds ratio (OR) for infertility. An OR > 1 denotes higher odds for infertility and OR < 1 denotes lower odds for infertility. OR represents the odds that an outcome happens given a particular exposure compared to the odds of an outcome happening without that exposure [31]. To test the linear trend of the FR and OR, P-trend was calculated using the quartiles in a continuous form.
Aside from analyzing the individual compounds, a mixture approach was also performed. Due to the high correlation among chemicals, survival analysis was not used to explore mixture effects. Instead, weighted quantile sum (WQS) regression, a strategy for estimating empirical weights for a weighted sum of quantiled concentrations most associated with the health outcome (TTP and binary infertility), was performed [32]. Chemicals of concern were identified with a threshold of 7.6% (100/number of chemicals). Bootstrapping was set to 100. The dataset was split into a 35% training set and a 65% holdout validation set.
Because interaction tests between chemicals and covariates showed significant results for age and use of COCs, we stratified the sample based on these two factors, namely age (with the cohort mean 29 years old as cut off) and use of COC (yes, no) as the most recent pre-pregnancy contraception method. Quartiles were specific for each age group to ensure equal n in each quartile. All models were adjusted for well-established risk factors such as parity, education, regularity of menses (regular, irregular), maternal body mass index (BMI), and smoking status based on serum cotinine concentrations as described above. We also tested other covariates such as education (lower secondary, upper secondary/vocational studies, university/college), physical activity (< 1 h/week, 1 h/week, 1–2 h/week, ≥3 h/week), and alcohol intake (never, seldom, once every other week, at least once a week). The only paternal covariate that was tested was BMI. However, none of these other covariates modified the FR and OR by 10% or more. Hence, they were not added to the model.
These analyses were performed using IBM SPSS Statistics, version 22.0 (IBM Corp., Armonk, NY, USA) and the R programming language [33] through RStudio [34] with the packages gWQS [35], ggplot2 [36], ggpubr [37], and reshape2 [38].