Maternal and paternal serum concentrations of persistent organic pollutants and the secondary sex ratio: A population-based preconception cohort study.

The secondary sex ratio (SSR) is the ratio of males to females at birth, while the primary sex ratio is the ratio at conception ( Buck Louis and Platt 2011 ). Across multiple academic disciplines including demography, sociobiology, epidemiology, and environmental science, the SSR has long been the subject of scientific investigation, serving as a useful tool in monitoring population dynamics and health ( McDonald et al. 2014 ). The stability and variability of the SSR observed at the population level have been proposed to be influenced by a variety of endogenous and exogenous factors, such as parental ages (Chahnazarian 1998; Jacobsen et al. 1999 ; Mathews and Hamilton 2005 ), birth order ( Biggar et al. 1999 ; Mathews and Hamilton 2005 ), race/ethnicity ( David et al. 2007 ; Mathews and Hamilton 2005 ), follicular phase length ( Martin 1997 ; Weinberg et al. 1995 ), timing of conception within the menstrual cycle ( Martin 1997 ; James 2008b ), stress ( Bae et al. 2017b ; Fukuda et al. 1998 ; Zorn et al. 2002 ), endocrine and immunological effects ( James 2008a ; Ober 1992 ), and other environmental factors ( Terrell et al. 2011 ). In particular, there has been speculation that recent declines in the SSR in some developed countries may be attributed to ubiquitous exposure to environmental toxicants ( Davis et al. 2007 ; Grech et al. 2003 ; Mathews and Hamilton 2005 ). Indeed, the decreasing trends in the SSR are juxtaposed with other adverse trends in reproductive outcomes, such as testicular cancer, genitourinary malformations, fecundity impairments, and gynecologic disorders ( Buck Louis et al. 2011a ; Skakkebeak et al. 2001 ). Of late, these phenomena have been synthesized in the paradigms of the testicular dysgenesis syndrome in males ( Skakkebeak et al. 2001 ) and the ovarian dysgenesis syndrome in females ( Buck Louis et al. 2011a ), providing conceptual frameworks for assessing environmental influences on human reproduction. Persistent organic pollutants (POPs) are of major public health concern, given their long half-lives and tendency to bioaccumulate and biomagnify in wildlife and humans. Exposure to POPs, such as dioxins, organochlorine pesticides (OCPs), polychlorinated biphenyls (PCBs), polybrominated biphenyls (PBBs), and polybrominated diphenyl ethers (PBDEs), particularly during the sensitive windows of human reproduction has been investigated in relation to a variety of reproductive outcomes, given the purported endocrine-disrupting properties of these chemicals ( Nicolopoulou-Stamati and Pitsos 2001 ; Toft et al. 2004 ; Vested et al. 2014 ). Along with considerable experimental evidence suggesting the harmful effects of POPs on reproduction in terms of oocyte maturation and follicle physiology ( Bhattacharya and Keating 2012 ; Pocar et al. 2003 ), a growing body of epidemiologic evidence has demonstrated that exposure to POPs may adversely affect human reproduction, possibly influencing reproductive hormones, menstrual cycles, semen quality, and couple fecundity among others ( Buck Louis et al. 2013 , 2016 ; Mumford et al. 2015 ; Nicolopoulou-Stamati and Pitsos 2001 ; Toft et al. 2004 ; Vested et al. 2014 ). Previous reports have suggested that parental exposure to POPs may be associated with alterations in the SSR. For instance, in landmark studies among the resident population in Seveso, Italy following a chemical manufacturing plant explosion in 1976, paternal exposure to high levels of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) was found to be associated with the reversal of the SSR (i.e., an excess of female births) ( Mocarelli et al. 1996 , 2000 ). Several studies have also investigated PCBs in relation to the SSR, demonstrating equivocal findings on the association between this chemical class and the SSR and possible opposing directions toward infant sex depending upon partners (i.e., paternal versus maternal exposure) or hormonal activities (i.e., estrogenic versus anti-estrogenic) of PCB congeners ( Nieminen et al. 2013 ; Taylor et al. 2007 ; Terrell et al. 2011 ). With regard to OCPs, the effects of exposure to dibromochloropropane (DBCP) ( Potashnik et al. 1984 ; Potashnik and Porath 1995 ), dichlorodiphenyltrichloroethane (DDT) ( Cocco et al. 2005 , 2006 ; Salazar-García et al. 2004 ), and hexachlorobenzene (HCB) ( Jarrell et al. 2002 ; Khanjani and Sim 2006 ) on the SSR have been evaluated in several studies, with paternal exposure to DBCP being somewhat consistently associated with the reversal of the SSR ( Potashnik et al. 1984 ; Potashnik and Porath 1995 ). However, there have been few studies assessing both paternal and maternal preconception exposure to POPs in relation to the SSR as an approach to ensuring the temporal order of offspring sex determination. To investigate the potential effects of preconception exposure to POPs on human sex selection, we designed the Longitudinal Investigation of Fertility and the Environment (LIFE) Study of multiple classes of POPs and the SSR. By design, a total of 56 POPs measured prior to conception in both male and female partners were assessed in relation to the SSR in accordance with the couple-dependent nature of human conception.

The LIFE Study is a prospective cohort study designed to investigate environmental influences on human fecundity and fertility, as previously described in detail ( Buck Louis et al. 2011b ). Briefly, 501 couples discontinuing contraception and trying for pregnancy were recruited from 16 counties in Michigan and Texas from 2005 to 2009, who were prospectively followed until pregnant or 12 months of attempting pregnancy. Women who became pregnant during the 12 months of follow-up were additionally followed to delivery or through a pregnancy loss. The eligibility criteria for participation were as follows: a) females aged 18–40 years and males aged 18 and older years; b) couples in a committed relationship; c) couples who were able to communicate in English or Spanish; d) females’ menstrual cycles ranging from 21 to 42 days; e) no use of injectable contraceptives within 12 months; and f) no history of physician-diagnosed infertility or sterilization procedures. Of the couples enrolled in the LIFE Study, 235 (46.9%) couples who had a singleton birth were used for the present study, excluding 2 (0.4%) couples who had multiple births and 264 (52.7%) couples without an observed live birth during the follow-up period. Institutional review board approvals were obtained from all collaborating institutions. Written informed consent was provided by all study participants prior to study participation. Baseline data were collected by research assistants in the couples’ home. Upon recruitment, the female partner provided a urine sample, which was used for a home pregnancy test to ensure that she was not pregnant. In-person interviews were conducted with each partner of the couple to ascertain socio-demographic characteristics (i.e., age, race/ethnicity, education, and annual household income) and reproductive history (i.e., parity and number of pregnancies fathered). Non-fasting blood (approximately 20 mL) was obtained from each partner of the couple for the quantification of serum concentrations of POPs and lipids. The blood samples were shipped on ice to the study’s laboratory and kept frozen at −20°C or colder until analysis. Couples who had a live birth during the follow-up period were asked to return standardized birth announcements following delivery to ascertain birth outcomes (i.e., infant sex, birth size, delivery mode, and date of birth). All analyses were conducted by the Division of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention. Specifically, couples’ serum concentrations (ng/g) were measured using isotope dilution high-resolution mass spectrometry for the following chemicals: a) 9 OCPs (HCB; β-hexachlorocyclohexane [β-HCH]; γ-hexachlorocyclohexane [γ-HCH]; oxychlordane; trans -nonachlor; o,p′-DDT; p,p′-DDT; p,p′-dichlorodiphenyldichloroethylene [p,p′-DDE]; and mirex); b) 1 PBB (PBB 153); c) 10 PBDEs (congeners #17, 28, 47, 66, 85, 99, 100, 153, 154, and 183); and d) 36 PCBs (congeners #28, 44, 49, 52, 66, 74, 87, 99, 101, 105, 110, 114, 118, 128, 138, 146, 149, 151, 153, 156, 157, 167, 170, 172, 177, 178, 180, 183, 187, 189, 194, 195, 196, 201, 206, and 209). Standard operating procedures were used for the quantification of a total of 56 serum POP concentrations, inclusive of ongoing quality assurance and control procedures ( Sjödin et al. 2004 ). The limits of detection (LODs) varied by analyte, ranging from 0.003 to 0.01 ng/g. All machine-observed values for serum chemical concentrations were utilized for analysis without automatic substitution of concentrations below the LODs or lipid adjustment to preclude bias associated with such practices ( Richardson and Ciampi 2003 ; Schisterman et al. 2006 ). Serum cotinine concentrations (ng/mL) were quantified using liquid chromatography-isotope dilution tandem mass spectrometry ( Bernert et al. 1997 ) to assess baseline exposure to smoking. Serum lipids (ng/g) were quantified using commercially available enzymatic methods ( Akins et al. 1989 ) and established calculation methods based on individual components including total cholesterol, free cholesterol, triglycerides, and phospholipids ( Phillips et al. 1989 ). In the descriptive phase of analysis, we assessed the distributions of all variables and inspected missing data and influential observations. Selected baseline characteristics of the study participants were evaluated by infant sex. Differences in continuous and categorical characteristics by infant sex were assessed using the nonparametric Wilcoxon test and Fisher’s exact test, respectively. Distributional properties of serum POP concentrations were summarized as geometric means (GMs) and 95% confidence intervals (CIs), along with selected percentiles (50th, 75th, 90th, 95th, and 99th). Differences in the GMs of serum POP concentrations by infant sex were assessed using the nonparametric Wilcoxon test. In the analytic phase, we used modified Poisson regression models ( Zou 2004 ) to estimate the relative risks (RRs) of a male live birth and corresponding 95% CIs for serum POP concentrations. In the models, serum POP concentrations were log-transformed to account for their skewed distribution and standardized by their standard deviation (SD) to aid in the interpretation of point estimates. Partner-specific models were run for maternal and paternal serum POP concentrations. Based on our review of the literature ( Biggar et al. 1999 ; Chahnazarian 1988 ; Fukuda et al. 2002 ; Jacobsen et al. 1999 ; Kolk and Schnettler 2015; Koshy et al. 2010 ; Mathews and Hamilton 2005 ), we adjusted a priori for serum lipid (ng/g; continuous), age (years; continuous), maternal parity (nulliparous/parous; for maternal serum POP concentrations only), annual income (< $70,000/≥ $70,000), serum cotinine (ng/mL; continuous), and the sum of remaining chemicals in the relevant class of compounds. In addition, couple-based models, which simultaneously included both partners’ serum POP concentrations, were run to control for potential confounding by the other partner’s exposure when assessing each partner’s exposure in relation to the risk of a male live birth. As far as the correlation between maternal and paternal serum POP concentrations is concerned, significant mild-to-moderate correlations (Spearman’s rank correlation coefficient range, 0.14–0.59) were observed for 53 POPs (except for PCB congeners #149, 178, and 195), serving as a motivation for the use of couple-based models. Given the exploratory nature of this work, statistical significance was initially denoted by p -value < 0.05. We subsequently assessed the significance at the 0.0009 (approximately equal to 0.05/56 [the number of POPs examined]) level to adjust for multiple comparisons. All statistical analyses were performed using SAS Version 9.4 (SAS Institute Inc., Cary, NC, USA).

Of the 235 live births, 116 (49.4%) were males and 119 (50.6%) were females, with the overall SSR of 0.97 (95% CI, 0.75–1.26) indicative of a female excess. The study cohort comprised predominantly non-Hispanic white and college-educated couples. The mean ages (± SD) of the female partners and the male partners were 29.8 (± 3.7) years and 31.5 (± 4.6) years, respectively. Approximately half of the female partners (53.2%) were parous and more than half of the male partners (57.5%) had previously fathered a pregnancy upon enrollment. Of the baseline characteristics examined, only one characteristic (i.e., paternal age) differed significantly by infant sex. Namely, the mean age (± SD) of men who fathered a boy (32.2 ± 5.1 years) was slightly higher than that of men who fathered a girl (30.8 ± 3.9 years; p -value, 0.03) ( Table 1 ). The distributions (GMs and 95% CIs) of serum POP concentrations by infant sex are presented in Table 2 . Selected percentiles (50th, 75th, 90th, 95th, and 99th) of serum POP concentrations are presented in Supplementary Table 1 . Most couples (≥ 99%) had serum POP concentrations above the LOD for 2 OCPs (i.e., HCB and p,p′-DDE), 3 PBDEs (i.e., congeners #47, 100, and 153), and 5 PCBs (i.e., congeners #74, 118, 138, 153, and 180). Of the 9 OCPs, p,p′-DDE had the highest GM concentrations (for female partners, 0.58 ng/g [95% CI, 0.53–0.63]; for male partners, 0.75 ng/g [95% CI, 0.70–0.81]). The PBDE congener with the highest GM concentrations was PBDE 47 (for female partners, 0.12 ng/g [95% CI, 0.10–0.13]; for male partners, 0.11 ng/g [95% CI, 0.10–0.13]). Of the 36 PCB congeners, PCB 153 had the highest GM concentrations (for female partners, 0.04 ng/g [95% CI, 0.04–0.05]; for male partners, 0.06 ng/g [95% CI, 0.05–0.06]). Serum POP concentrations differed significantly by infant sex for paternal p,p′-DDE, 2 paternal PBDE congeners (i.e., congeners #154 and 183), and 4 maternal PCB congeners (i.e., congeners #87, 156, 157, and 167) with slightly higher serum concentrations in boys than in girls before rounding to two or three decimal places, except for maternal PCB 157 ( Table 2 ). Table 3 presents the RRs of a male birth by maternal and paternal serum POP concentrations, only for POPs with any significant associations with a male birth across the models. See Supplementary Table 2 for corresponding estimates for all other POPs examined. When couples’ serum POP concentrations were modeled jointly in combination with other covariates, significant associations between select POPs and the SSR were noted. Specifically, of the 56 POPs examined, maternal PCB 128 and paternal HCB were significantly associated with an excess of female births (adjusted RRs, 0.75 [95% CI, 0.60–0.94] and 0.81 [95% CI, 0.68–0.97] per 1 SD increase in log-transformed serum chemical concentrations, respectively). On the other hand, maternal mirex, paternal PCB 128, and paternal p,p′-DDE were significantly associated with an excess of male births (adjusted RR range, 1.10–1.22 per 1 SD increase in log-transformed serum chemical concentrations). However, when the significance was assessed at the 0.0009 level, only maternal mirex remained significantly associated with the SSR ( Table 3 ).

In this population-based prospective cohort study, we found no conclusive evidence supporting an association between parental preconception exposure to POPs and the SSR. In fact, in our exploratory analysis of multiple classes of POPs, 2 serum POP concentrations in mothers (i.e., mirex and PCB 128) and 3 (i.e., HCB, p,p′-DDE, and PCB 128) in fathers were significantly associated with an excess of male or female births, when couples’ serum POP concentrations were modeled jointly in combination with other covariates. However, of note is the lack of robustness in the significant findings depending upon model specification or statistical methods used, possibly indicating uncertain associations of these select POPs with the SSR. Furthermore, when we adjusted for multiple comparisons by assessing the significance at the 0.0009 level, only the association observed for maternal serum mirex concentrations and an excess of male births remained significant. When the tertiles of serum POP concentrations were used in the couple-based multivariate-adjusted model, a possible dose-response relation was noted for maternal mirex (2nd versus 1st tertile, adjusted RR, 1.06 [95% CI, 0.74–1.52]; 3rd versus 1st tertile, adjusted RR, 1.33 [95% CI, 0.91–1.92]; data not shown), although not significant. Still, it is worth noting that serum mirex concentrations were below the LOD in 82% of the female partners. Meanwhile, serum mirex concentrations in the U.S. general population from the National Health and Nutrition Examination Survey (NHANES) were generally below the LOD (somewhere between 75% and 90% of the population) during the comparable time period ( Centers for Disease Control and Prevention 2017 ). As far as we know from the literature, our study is the first report on the association between maternal serum mirex concentrations and the SSR, underscoring the need for corroboration through further research on this topic. Our study appears to be in line with some but not all previous reports suggesting significant associations between select POPs and the SSR. In a recent review of more than 100 studies on environmental and occupational toxicants and the SSR, Terrell et al. (2011) concluded that paternal exposure to dioxins may be associated with a decreased SSR, whereas paternal exposure to PCBs may be associated with an increased SSR. On the other hand, in a systematic review of 15 studies on PCBs and the SSR, Nieminen et al. (2013) found no strong or moderate indication that parental exposure to PCBs altered the SSR from the historical reference range. In fact, a small number of studies have suggested that maternal serum PCB concentrations may be associated with a decreased SSR ( Hertz-Picciotto et al. 2008 ; Weisskopf et al. 2003 ). In a retrospective cohort of 173 mothers with a live birth from the Great Lakes region of the United States between 1970 and 1995, the adjusted odds ratio (OR) for a male birth among mothers in the highest quintile of serum PCB concentrations was 0.18 (95% CI, 0.06–0.59) compared with mothers in the lowest quintile ( Weisskopf et al. 2003 ). A prospective cohort study of 339 pregnant women in the San Francisco Bay Area during the 1960s showed that the RR of a male birth decreased by 33% comparing women at the 90th percentile of a total of 11 PCBs with women at the 10th percentile (RR, 0.67; 95% CI, 0.48–0.94) ( Hertz-Picciotto et al. 2008 ). Meanwhile, a preliminary study on maternal preconception exposure to PCBs and the SSR was suggestive of varying effects of PCB congeners on the SSR depending upon their hormonal activities ( Taylor et al. 2007 ). Namely, in a cohort of 50 women with a live birth from the New York State Angler Cohort Study, maternal preconception exposure to estrogenic PCBs (congeners #4, 8, 15, 18, 31, 44, 47, 48, 52, 70, 77, 99, 101, 126, 136, 153, and 188) but not anti-estrogenic PCBs (congeners #77, 105, 114, 126, 156, and 169) was associated with an increased SSR, although the association did not reach statistical significance ( Taylor et al. 2007 ). In the present study, we found that only one out of the 36 PCB congeners examined (i.e., PCB 128 as a non-dioxin-like PCB) was significantly associated with the SSR in the couple-based multivariate-adjusted models, demonstrating opposing associations of this chemical with the SSR depending upon partners. Specifically, paternal PCB 128 was associated with an excess of male births, whereas maternal PCB 128 was associated with an excess of female births. However, it is noteworthy that serum PCB 128 concentrations were relatively low and below the LOD in 98% of the female partners and in 99% of the male partners. Likewise, serum PCB 128 concentrations in the U.S. general population from the NHANES were mostly below the LOD during the comparable time period ( Centers for Disease Control and Prevention 2017 ). Furthermore, when we adjusted for multiple comparisons, the associations of maternal and paternal PCB 128 with the SSR were no longer significant ( p -value ≥ 0.0009), indicative of the possibility of chance findings. Although we did not examine parental exposure to dioxins, we did assess the effects of dioxin-like PCBs (i.e., mono-ortho-substituted PCBs; congeners #105, 114, 118, 156, 157, 167, and 189) on the SSR, without noticing any significant associations. Of note, although not significant, some maternal estrogenic PCBs (e.g., congeners #44 and 52) were associated with an increased SSR (in the couple-based models, adjusted RRs, 1.86 [95% CI, 0.49–7.09] and 1.64 [95% CI, 0.45–6.06] per 1 SD increase in log-transformed serum chemical concentrations, respectively; Table 3 ), consistent with a previous report on maternal estrogenic PCBs and the SSR ( Taylor et al. 2007 ). Although our study is the first investigation known to us on the association between mirex and the SSR, environmental or occupational exposure to other OCPs such as HCB and DDT has been assessed in relation to the SSR in some previous studies. For instance, exposure to HCB was not found to be associated with alterations in the SSR in a study using the Turkish national data from 1935 to 1990 ( Jarrell et al. 2002 ) or in a cohort of primiparous women enrolled in the 1990s in Austria ( Khanjani and Sim 2006 ). In addition, exposure to DDT in the anti-malaria campaign in some countries such as Italy ( Cocco et al. 2005 , 2006 ) and Mexico ( Salazar-García et al. 2004 ) has been investigated in relation to the SSR, demonstrating equivocal findings on the effects of exposure to DDT (as measured by estimated concentrations of p,p′-DDE [the most persistent metabolite of DDT] in fat) on the SSR. Meanwhile, our study showed that paternal serum HCB concentrations were significantly associated with a decreased SSR (in the couple-based model, adjusted RR, 0.81 [95% CI, 0.68–0.97] per 1 SD increase in log-transformed serum chemical concentrations; p -value ≥ 0.0009; Table 3 ), consistent with some previous studies indicating male-mediated effects of select chemicals (e.g., TCDD and DBCP) on the reversal of the SSR ( Mocarelli et al. 1996 , 2000 ; Potashnik et al. 1984 ; Potashnik and Porath 1995 ). Contrarily, paternal serum concentrations of p,p′-DDE were found to be significantly associated with an increased SSR (in the couple-based model, adjusted RR, 1.22 [95% CI, 1.03–1.44] per 1 SD increase in log-transformed serum chemical concentrations; p -value ≥ 0.0009; Table 3 ) in our study, although we cannot rule out the possibility of a chance finding. In the present study, no statistically significant associations of PBB 153 and PBDEs with the SSR were observed. While some maternal PBDEs (i.e., congeners #28, 99, and 154) showed significant associations with an excess of male births in the couple-based crude models (RR range, 1.09–1.12 per 1 SD increase in log-transformed serum chemical concentrations; Table 3 ), the associations no longer remained significant after the adjustment for covariates. When we further evaluated the effects of covariates on the associations between these PBDEs and the SSR in the couple-based multivariate-adjusted models, only delta (paternal age − maternal age) was found to be significantly associated with the risk of a male birth (adjusted RR, 1.04; 95% CI, 1.01–1.08; data not shown). To the best of our knowledge, there have been no previous studies investigating the association between serum PBDE concentrations and the SSR, underscoring the need for additional research on this topic. As for PBBs, in a cohort of 300 couples from the Michigan Long-Term PBB Study, combined parental exposure to PBBs increased the odds of a male birth, although the finding did not reach statistical significance ( Terrell et al. 2009 ). Meanwhile, perfluoroalkyl and polyfluoroalkyl substances (PFASs), an emerging class of POPs, have been first evaluated in relation to the SSR in our previous study using data from the LIFE Study, suggesting paternal exposure to select PFASs (i.e., paternal N -methyl-perfluorooctane sulfonamidoacetic acid and perfluorononanoic acid) being associated with the reversal of the SSR ( Bae et al. 2015 ). While research efforts have been made to identify genetic (e.g., the SRY [sex-determining region Y] gene) and environmental determinants of offspring sex, the mechanistic underpinnings of human sex selection are still obscure. It has been thought that multiple factors in fathers (e.g., sperm Y:X chromosome ratio and intrinsic differences in size and swimming velocity between Y- and X-chromosome bearing sperm) and mothers (e.g., sperm selection within the female reproductive tract and differential survival of male and female embryos) may contribute to offspring sex determination ( Almiñana et al. 2014 ; Davis et al. 2007 ; Jongbloet 2004 ; Tiido et al. 2005 ). One of the prevailing hypotheses on the SSR is the James’ hormonal hypothesis, which theorizes that parental hormone levels around the time of conception are partly responsible for alterations in the SSR ( James 2008a , 2008b , 2012 , 2013 ). According to this hypothesis, high levels of gonadal hormones, such as testosterone (of either parent) and estrogen (of mother), are associated with an excess of male births, whereas high levels of gonadotropins, such as follicle-stimulating hormone (FSH) and luteinizing hormone (LH), of either parent are associated with an excess of female births ( James 2013 ). Given the purported endocrine-disrupting effects of POPs, some of our findings (e.g., maternal estrogenic PCBs and an increased SSR [although not significant]; paternal HCB [ Ralph et al. 2003 ] and a decreased SSR) appear to be congruent with the hormonal hypothesis, whereas others (e.g., paternal p,p′-DDE [ Ayotte et al. 2001 ; Martin et al. 2002 ] and an increased SSR) do not. Existing evidence on the hormonal activities of mirex is conflicting, making it somewhat challenging to interpret our findings on maternal serum mirex concentrations and an increased SSR based on the hormonal hypothesis. For instance, when tested by the E-SCREEN assay, mirex did not show estrogenic activity without inducing the proliferation of MCF-7 human breast cancer cells ( Soto et al. 1995 ). In an androgen-responsive reporter gene assay, mirex did not antagonize the effect of 5α-dihydrotestosterone, exhibiting no interaction with the androgen receptor ( Schrader and Cooke 2000 ). Along with its questionable in vitro estrogenic or anti-androgenic effects, mirex was not found to be significantly associated with the levels of sex hormones such as estradiol, FSH, and LH among 210 premenopausal women, although an inverse association between LH and serum mirex concentrations was noted among 77 peri- and postmenopausal women in a cross-sectional study conducted in Brazil ( Freire et al. 2014 ). On the other hand, in a population-based case-control study conducted in the United States, serum mirex concentrations were associated with an increased risk of endometriosis, an estrogen-dependent gynecologic disorder ( Upson et al. 2013 ). In a nested case-control study of 48 newborns with diagnosis of cryptorchidism and/or hypospadias and 114 controls, mirex measured in placenta tissues was significantly associated with an increased risk of genitourinary malformations, suggesting that exposure to mirex may provoke an unbalanced androgen/estrogen ratio and lead to abnormal male sexual differentiation ( Fernandez et al. 2007 ; Skakkebaek et al. 2001 ). As such, it is difficult to interpret the observed findings on POPs and the SSR simply based on the hormonal hypothesis, partly because these chemicals may have more than one mode of action (e.g., both estrogenic and anti-androgenic) and exhibit different modes of action at different doses. Although we assessed each chemical in relation to the SSR while adjusting for the sum of remaining chemicals in the relevant class of compounds, it is important to keep in mind that chemical mixtures may act together to disturb the endocrine system. In addition, the background levels of endogenous hormones in males and females may play a role in the mode(s) of action of a given chemical, possibly explaining the discrepancy in the findings on POPs and the SSR by partner. In consideration of toxicokinetics and toxicodynamics, it is also worth noting that different metabolites may have different hormonal activities compared with their parent compound ( Andersson et al. 2016 ). Together with taking all these factors that may influence the hormonal milieu into account, possible non-hormonal mechanisms by which exposure to POPs alters the SSR (e.g., sperm abnormalities) deserve more research attention ( Bae et al. 2017a ; Fukuda et al. 1998 ). The strengths of the present study include the quantification of serum POP concentrations prior to conception, along with the prospective cohort design. Our couple-based cohort enabled us to incorporate both partners’ serum POP concentrations into the same models when assessing a couple-dependent birth outcome. Given that POPs may have different endocrine-disrupting effects for men and women, it is necessary to assess each partner’s possible role in determining offspring sex while controlling for the other partner’s. However, the major limitations of the present study should be considered when interpreting the study results. We were unable to evaluate the primary sex ratio or sex ratio at conception due to the difficulty in measuring conception at the population level. With the absence of data on preconception hormone levels, we were unable to fully explore the purported hormonal activities of POPs and their associations with the SSR. The possibility of residual confounding or model misspecification should be taken into account, given the uncertainty as to factors influencing the SSR. Although we believe that our cohort is one of the largest couple-based cohorts with preconception recruitment, its sample size may be relatively small with regard to the detection of variability in the SSR. As we undertook multiple tests to assess a number of POPs in relation to the SSR, the possibility of chance findings cannot be ruled out. Given the sampling of couples with planned pregnancy, the study results may not be generalizable to couples with unplanned pregnancy or the general population.

In a comprehensive evaluation of multiple classes of POPs in relation to the SSR, we found that specific POPs (i.e., mirex and PCB 128 in mothers; HCB, p,p′-DDE, and PCB 128 in fathers) were significantly associated with alterations in the SSR, consistent with some but not all earlier findings on POPs and the SSR. However, the significant findings noted in our study await corroboration, given the absence of previous investigation on certain POPs (e.g., mirex and PBDEs) and the inconsistencies over the effects of each chemical on offspring sex determination. Furthermore, our findings need to be interpreted with caution, given our exploratory analytic approach and major limitations of the study. With the lack of conclusive evidence on the association between parental preconception exposure to POPs and the SSR, we wish to stress the need for further confirmatory research particularly focusing on the mechanistic underpinnings of human sex selection. Preferably, larger couple-based preconception cohort studies, which allow the measurement of time-sensitive exposures for couples, are warranted to elucidate the effects of POPs on human sex selection, as a parentally-mediated reproductive event. Such research efforts will facilitate understanding of the potential health implications of exposure to these ubiquitous chemicals at environmentally relevant doses.