Association of Hexachlorobenzene (HCB), Dichlorodiphenyltrichloroethane (DDT), and Dichlorodiphenyldichloroethylene (DDE) with in Vitro Fertilization (IVF) Outcomes

Background: Hexachlorobenzene (HCB), dichlorodiphenyltrichloroethane (DDT), and dichlorodiphenyldichloroethylene (DDE) are persistent chlorinated pesticides with endocrine activity that may adversely affect the early stages of human reproduction. Objective: Our goal was to determine the association of serum levels of HCB, DDT, and DDE with implantation failure, chemical pregnancy, and spontaneous abortion in women undergoing in vitro fertilization (IVF) from 1994 to 2003. Methods: Levels of HCB and congeners of DDT and DDE were measured in serum collected during the follicular phase. Multivariable-adjusted statistical models accommodating multiple outcomes and multiple cycles per woman were used to estimate the relation between serum pesticide levels and IVF outcomes. Results: A total of 720 women with a mean ± SD age 35.4 ± 4.2 years at enrollment contributed 774 IVF cycles. All samples had detectable levels of HCB, DDT, and DDE, with median levels of 0.087 ng/g serum for HCB, 1.12 ng/g serum for total DDT, and 1.04 ng/g serum for p,p´-DDE. Compared with the lowest quartile (Q1) of HCB, the lipid- and multivariable-adjusted odds ratio (OR) for failed implantation was significantly elevated for those with higher HCB quartiles [Q2–Q4; adjusted ORs: for Q2, 1.71; 95% confidence interval (CI): 1.03, 2.82; for Q3, 2.30; 95% CI: 1.39, 3.81; for Q4, 2.32; 95% CI: 1.38, 3.90] and showed a significantly increasing trend (p = 0.001). No statistically significant associations were observed between DDT/DDE and IVF outcomes or between HCB and chemical pregnancy or spontaneous abortion. Conclusions: Serum HCB concentrations were on average lower than that of the general U.S. population and associated with failed implantation among women undergoing IVF.

volume 120 | number 2 | February 2012 • Environmental Health Perspectives Research | Children's Health Hexachlorobenzene (HCB) and dichloro diphenyltrichloroethane (DDT) are organo chlorines (OCs) used in fungus and pest control. These OCs have demonstrated estro genic activity in in vitro assays and adverse reproductive effects in fish and wildlife (Tiemann 2008). Although HCB and DDT have been banned in the United States, HCB is a byproduct of organic chemical produc tion processes and is still produced outside the United States [Agency for Toxic Substances and Disease Registry (ATSDR) 2002b]. DDT continues to be used abroad for the control of mosquitoborne diseases (van den Berg 2009).
The main source of human exposure to these OCs is through dietary ingestion. Generally, HCB and DDT are found in meat, fish, and milk products (Brilhante and Franco 2006;Fontcuberta et al. 2008;Mawussi et al. 2009;Yu et al. 2009). DDT can be found in higher concentrations in foods imported from countries still using DDT (ATSDR 2002a).
Because of their chemical stability, bioaccu mulation up the food chain, and the ongoing production or use in some countries, HCB, DDT, and dichlorodiphenyldichloroethyl ene (DDE), the primary metabolite of DDT, continue to be detected in human blood (ATSDR 2002a), breast milk (Tsang et al. 2011), follicular fluid (Meeker et al. 2009), amniotic fluid (Foster et al. 2000;Luzardo et al. 2009), and human umbilical cord blood (Jimenez Torres et al. 2006).
There are conflicting data regarding the effect of exposure to DDT/DDE on early reproductive outcomes. A recent study described no association between oocyte, fer tilization, and implantation parameters in women undergoing in vitro fertilization (IVF) and exposure to DDT (AlSaleh et al. 2009). Another study found an association between increased time to pregnancy and selfreport of exposure to agricultural and home pesticides (Harley et al. 2008). Some studies suggest an association of DDT/DDE with early clinically detected fetal loss (Longnecker et al. 2005;Venners et al. 2005). Other studies reported early pregnancy loss and early neonatal and childhood mortality after a widespread acciden tal exposure to HCB in Turkey (Jarrell et al. 1998;Peters et al. 1982). Because of concerns raised by these earlier studies, we explored the association of serum levels of these persistent OCs with early reproductive failure uniquely observable within the context of an IVF study.

Study population.
This study population is a nested case-control subset of patients derived from a prospective cohort designed to assess predictors of IVF outcomes. Between August 1994 and June 2003, couples commencing assisted reproduction therapy were enrolled through three Boston, Massachusetts, area clinics. Couples agreeing to participate in this study underwent an informed consent process and provided written consent before enrollment. There were two phases of patient enrollment, 1994-1999 and 2000-2003, cor responding to initial funding and renewal phases. Study protocols were approved by the Institutional Review Board at Brigham and Women's Hospital and the Human Subjects Committee at Harvard School of Public Health. Approximately 65% of couples who were approached agreed to participate in the study. Couples who required donor gametes, Background: Hexachlorobenzene (HCB), dichlorodiphenyltrichloroethane (DDT), and dichloro diphenyldichloroethylene (DDE) are persistent chlorinated pesticides with endocrine activity that may adversely affect the early stages of human reproduction. oBjective: Our goal was to determine the association of serum levels of HCB, DDT, and DDE with implantation failure, chemical pregnancy, and spontaneous abortion in women undergoing in vitro fertilization (IVF) from 1994 to 2003. Methods: Levels of HCB and congeners of DDT and DDE were measured in serum collected during the follicular phase. Multivariableadjusted statistical models accommodating multiple out comes and multiple cycles per woman were used to estimate the relation between serum pesticide levels and IVF outcomes. results: A total of 720 women with a mean ± SD age 35.4 ± 4.2 years at enrollment contributed 774 IVF cycles. All samples had detectable levels of HCB, DDT, and DDE, with median levels of 0.087 ng/g serum for HCB, 1.12 ng/g serum for total DDT, and 1.04 ng/g serum for p,p´DDE. Compared with the lowest quartile (Q1) of HCB, the lipid and multivariableadjusted odds ratio (OR) for failed implantation was significantly elevated for those with higher HCB quartiles [Q2-Q4; adjusted ORs: for Q2, 1.71; 95% confidence interval (CI): 1.03, 2.82; for Q3, 2.30; 95% CI: 1.39, 3.81; for Q4, 2.32; 95% CI: 1.38, 3.90] and showed a significantly increasing trend (p = 0.001). No statistically significant associations were observed between DDT/DDE and IVF outcomes or between HCB and chemical pregnancy or spontaneous abortion. conclusions: Serum HCB concentrations were on average lower than that of the general U.S. population and associated with failed implantation among women undergoing IVF. key words: assisted reproduction, dichlorodiphenyldichloroethylene, dichlorodiphenyl trichloroethane, hexachlorobenzene, prospective cohort. Environ Health Perspect 120:316-320 (2012). http://dx.doi.org/10.1289/ehp.1103696 [Online 29 August 2011] a gestational carrier, or gamete/zygote intr afallopian transfer were excluded from the main study. Most couples underwent mul tiple IVF cycles (up to six), with an average of two cycles per couple.
Because of budget limitations, we were unable to measure serum OC levels on all of the study participants. Therefore, we devised our case and control selection strategy to max imize statistical power, selecting cases based on outcomes of interest. Of the 2,350 couples in the main study, 765 women were selected by the outcome of their first IVF cycle (failed implantation, chemical pregnancy, sponta neous abortion, or live birth). Because the number of available cases varied by failure type, with failed implantation being most fre quent, we analyzed serum HCB/DDT/DDE concentrations for a random sample of 200 women with firstcycle implantation failures. Because chemical pregnancies and spontane ous abortions were less frequent, all cycles from the entire cohort with these outcomes were selected. Also included in the analysis were 62 repeated cycles among a subset of 46 women, from a total of 265 cycles from 110 women who were selected for measurement of DDE, DDT, and HCB in a prior analysis of withinwoman variability in serum concentra tions of OC compounds from the same parent study (Meeker et al. 2009). A corresponding sample of firstcycle live births was selected using stratified sampling based on age cat egory, study center, and study phase. In total, 765 women contributed 827 cycles.
Data abstracted from clinical records were infertility diagnoses, ovarian stimula tion protocol (downregulation or flare), type and total ampules of gonadotropin, use of intracytoplasmic sperm injection (ICSI), day of embryo transfer (ET), number of embryos transferred, and cycle outcomes.
Only women who proceeded to undergo ET of one or more embryos were eligible for the present analysis. When at least one embryo was transferred but without subsequent rise of human chorionic gonadotropin (hCG) beyond 5.0 mIU/mL, the cycle outcome was defined as a failure of implantation. A chemical preg nancy was defined by a postET hCG meas urement of ≥ 5.0 mIU/mL with no further embryonic development (gestational sac or yolk sac). Clinical pregnancy was determined by ultrasound visualization of a gestational sac or a fetal heartbeat. Among clinically recognized pregnancies, outcomes included spontaneous abortion (fetal demise before 20 weeks of gesta tion) or live birth of at least one infant. Other outcomes for clinically recognized pregnancy were ectopic pregnancy, molar pregnancy, and stillbirth (fetal demise at ≥ 20 gestational weeks) and were excluded in the present anal ysis because numbers were small and lacked power for independent investigation.

Measurement of HCB, DDT, and DDE in serum.
Blood samples were collected from women during the follicular phase of their first IVF/ICSI cycle and subsequent cycles in a non fasting state immediately before hCG adminis tration. The serum fraction was separated for all blood samples by centrifugation and stored at -80°C. Measurement of the OC pesticides HCB, p,p´DDT, p,p´DDE, o,p´DDT, and o,p´DDE was conducted by the Organic Chemistry Analytical Laboratory, Harvard School of Public Health (Boston, MA), using methods described previously (Korrick et al. 2000). Briefly, after liquid-liquid extraction and column chromatography, samples were analyzed by dual gas chromatography with electron capture detection, on two capillary columns of different polarity using two inter nal standards. Samples were accompanied by the following quality control samples: a pro cedural blank, matrix spike samples, and a laboratory control sample. Each sample was spiked with two surrogate compounds to mon itor the efficiency of the extraction procedure. All final results were reported after subtract ing the amount of the analyte measured in the procedural blank associated with the ana lytic batch. Method detection limits (MDLs) were determined as recommended by the U.S. Environmental Protection Agency (1984).
The wetweight MDL values for HCB and congeners of DDT and DDE were all < 0.01 ng/g. The OC concentrations were measured at part per billion concentrations. The withinassay variation for all target pesti cides ranged from 2% to 5%. The between assay variation (percent coefficient of variation) for all target pesticides was < 10%.
Serum total cholesterol and triglycerides were measured enzymatically. Total serum lipid content was calculated by Equation 2 of Phillips et al. (1989) and then used for lipid standardization of serum OC concentration.
Statistical analysis. Serum OC concentra tions in the study population of 765 women were grouped into quartiles (Q1-Q4). Total DDT was defined for this analysis as a sum of p, p´DDT, o,p´DDT, p,p´DDE, and o,p´DDE. The relation between IVF out comes and quartiles of HCB, p,p´DDE, and total DDT was explored with multivariable generalized linear regression models, with data structured to accommodate joint models for multiple outcomes and multiple cycles con tributed per woman. Whenever a woman was at risk of a cycle failure because of implanta tion failure, chemical pregnancy, or spontane ous abortion in each cycle, a binary outcome (1 = failure, 0 = not a failure) was recorded. Thus, a woman provided up to three contri butions to risk sets for the possible outcomes within each IVF cycle. A logistic regression model was then employed for each early preg nancy outcome, while adjusting for covariates.
Interaction terms between OC quartile and failure type were included in the model to allow for a distinct relation between the OC of interest and reproductive failure type. Interaction terms between age and failure type were also included to account for the possibil ity of differing age effects on the various failure end points. A womanspecific randomeffect term, assumed to be Gaussian, was included to account for the correlation among outcomes from different cycles for the same woman. Trend tests were conducted by assigning ordi nal integer values to HCB, DDE, and DDT quartiles (0 = Q1 to 3 = Q4).
Potential confounding variables consid ered in our analyses were maternal age (< 35, 35-37, 38-40, > 40 years), body mass index (BMI), serum lipids, smoking status (never, past, or current), site (of the three participating Boston area clinics), study phase (1994-1998 or 1999-2003), race (Caucasian or other), pre vious live birth (yes/no), ampules of gonado tropins, protocol (downregulation or other), ICSI (yes/no), number of embryos transferred, and primary infertility diagnosis. Primary infer tility diagnoses were categorized as male factor, ovulatory dysfunction, and other/unexplained. The "other/unexplained" category included Müllerian anomalies and uterine or tubal pathology. Cycles that were missing race, infer tility diagnosis, or previous live birth (6% of cycles) were assigned a missing indicator vari able and retained in all analyses. Cycles missing any other covariate information were excluded.
The serum OC concentrations were stan dardized by dividing by the calculated total serum lipid concentration. In addition, we also modeled IVF outcomes in relation to wetweight serum levels of the specific and adjusted for serum lipids as a covariate in multivariable models, to reduce bias, as sug gested by Schisterman et al. (2005).

Results
Of the 827 IVF cycles evaluated, 774 (94%) had sufficient covariate information to be included in our analysis, representing 720 (94%) of the 765 women. Among the 54 IVF cycles excluded, 6 were missing BMI, 13 were missing lipids, 2 were missing ampules of gonadotropins, and 32 were missing num ber of embryos transferred. The 720 women had a mean age of 35 years, were primarily Caucasian (90%) and current non smokers (94%), and had a mean BMI of 24.2 kg/m 2 (Table 1). Unexplained infertility (34%) and male factor (35%) accounted for most infer tility diagnoses; the remaining women were diagnosed with either tubal factor (18%) or ovulatory dysfunction (13%). Because of volume 120 | number 2 | February 2012 • Environmental Health Perspectives the nested case-control sampling design, the oversampling of failure outcomes on a subset of women was not representative of the failure outcome percentages of the main study. There were 541 total cycle failures, of which 490 were included in the analysis and consisted of implantation failure (209), chemical preg nancy (161), and spontaneous abortion (120).
Distributions of serum concentrations of HCB, DDT, and DDE are presented as both wetweight and lipidstandardized con centrations (Table 2). No samples were below the limit of detection. The geometric mean (GM) and median concentration for HCB were 0.088 and 0.087 ng/g serum, respec tively, compared with a GM and median of 0.095 and 0.092 ng/g serum in the gen eral U.S. female population in [2003][2004] [Centers for Disease Control and Prevention (CDC) 2009]. Total DDT had the highest concentration (median = 1.12 ng/g serum, range 0.02-31.8 ng/g serum). Most of the total DDT was from p,p´DDE contributions, which accounted for > 90% of the sum of all measured DDT/DDE congeners. The GM and median p,p´DDE concentrations were, respec tively, 1.09 and 1.04 ng/g serum, with a range of 0.007-30.6 ng/g serum, compared with a GM and median of 1.45 and 1.25 ng/g serum in the general female U.S. population (CDC 2009). The GM and median for p,p´DDT in the general female U.S. population included levels below the limit of detection (CDC 2009). There was a strong correlation between the concentration of total DDT and p,p´DDE (Spearman correlation coefficient = 0.98). Of the 46 women with serum from multiple cycles ultimately included in the final analysis, 29 (63%) remained in the same HCB quartile across cycles after lipid standardization, simi lar to the 34 (71%) with lipid adjustment in the multivariable model. HCB concentrations between cycles were also strongly correlated (Spearman correlation coefficient = 0.98).
When serum lipids were adjusted as a covariate with other potential confounders, a dose-response relation between increasing HCB concentration quartile and odds ratios (ORs) for failed implantation was observed (Table 3). Compared with women in Q1, the multivariable adjusted ORs for failed implantation were 1.71 for women in Q2 [95% confidence interval (CI): 1.03, 2.82], 2.30 for women in Q3 (95% CI: 1.39, 3.81), and 2.32 for women in Q4 (95% CI: 1.38, 3.90; pvalue, test for linear trend = 0.001). When lipidstandardized HCB concentrations were used as the dependent variable, a statisti cally significant linear increase in ORs for implantation failure remained. Although no statistically significant associations were found between serum HCB and chemical pregnancy, spontaneous abortion, or early pregnancy fail ure, ORs for early pregnancy failure increased with increasing HCB quartiles. Lack of statis tical significance may be due to the relatively small numbers of chemical pregnancies and spontaneous abortions.
Although no statistically significant asso ciations were found between p,p´DDE and any individual or combined early reproduc tive failures, the OR for any of the three early pregnancy failure types among cycles in Q4 p,p´DDE versus Q1 was 1.6 (95% CI: 0.98, 2.63) ( Table 4). The OR for Q4 total DDT, after lipid adjustment, and early pregnancy failure versus Q1 was 1.85 (95% CI: 1.13, 3.03; pvalue, test for linear trend = 0.049; data not shown).
We additionally evaluated the relations of serum wetweight concentrations of DDE, DDT, and HCB, without lipid adjustment or lipid standardization, and outcomes, and effect estimates were altered negligibly (data not shown).

Discussion
Among the 765 women undergoing IVF in this study, all had detectable serum levels of HCB, DDT, and DDE. Implantation failure ORs exceeded 2 for women with serum con centrations of HCB in Q4, compared with women in the lowest quartile. Previous studies within this cohort have shown that serum OC concentrations were strongly correlated with follicular fluid concentrations (rvalues: HCB, 0.66; p,p´DDE, 0.92; p,p´DDT, 0.74) (Meeker et al. 2009), and in other popula tions OCs were detectable in amniotic fluid (Foster et al. 2000), suggesting the possibility of direct gamete and embryo exposure.
We are not aware of another study that describes the association between clinically observed implantation failure and exposure to HCB; therefore, it will be critical that further studies by other investigators replicate this finding. However, a non human primate study demonstrated that HCB exposure induced a dosedependent suppression of luteal serum progesterone (Foster et al. 1992). Suppressed luteal progesterone may increase the likeli hood of implantation failure due to reduced Values are mean ± SD or n (%). a Cycle-level data. b Primary infertility diagnosis. endometrial maturation or of early pregnancy loss from inadequately low levels of proges terone production (Jones and Wentz 1976).
Other studies demonstrated that HCB expo sure was associated with degenerative changes in ovarian structure, including germ cell destruction in rhesus monkeys (Iatropoulos et al. 1976) and primordial germ cell loss in cynomolgus monkeys (Jarrell et al. 1993), suggesting other potential mechanism of impaired reproductive outcome. Our study did not find a statistically significant association between HCB and chemical pregnancy or spontaneous abortion. However, exposure to HCBcontaminated foods in Turkey between 1955 and 1957 resulted in detectable HCB in serum drawn 40 years later. In that population, higher HCB levels were associated with a self reported his tory of spontaneous abortion (Jarrell et al. 1998). Furthermore, time to pregnancy, a possible surrogate marker for implantation failure, was not assessed in that study.
In the present study, we also found no significant association between serum DDT/ DDE and implantation failure, chemical preg nancy, or early pregnancy loss. However, we did observe a non significant OR of 1.6 for any of the three early pregnancy failure types among cycles in p,p´DDE Q4 compared with those in Q1.
One study that evaluated serum DDE lev els among 1,717 women and assessed past preg nancy outcomes via survey found a significant adjusted OR of 1.4 (95% CI: 1.1, 1.6) for fetal loss in women per 60µg/L increase of serum DDE (Longnecker et al. 2005). The primary limitation of that study was the retrospective design in which current serum DDE concen trations were associated with past fetal loss. Another study that followed newly married female textile workers in China also found an association between serum total DDT (range, 5.5-113.3 ng/g serum; median, 27.9 ng/g) and subsequent early fetal loss, defined as < 6 weeks of gestation as measured through decline in urinary hCG (Venners et al. 2005). In a smaller study with 99 women undergoing their first cycle of IVF, no statistically significant relationships between serum and follicular con centrations of DDT/DDE and oocyte num ber, quality, fertilization rates, or pregnancy rates were reported, although serum DDT/ DDE concentrations were higher than in our study population. The authors suggested that some "statistically suspicious" associations were observed; however, ORs, CIs, and pvalues were not given (Jirsova et al. 2010).
One limitation of the present study is that the study population consisted of cou ples undergoing IVF. It is unknown whether infertile couples respond differently to OC exposure relative to fertile couples or women with untested fertility. Therefore, although  the results of this study are internally valid, it is unclear whether the associations of between HCB or DDE/DDT and early pregnancy out comes would differ among women who have never experienced infertility. However, in the spontaneous conception setting, it would not be possible to evaluate the association between these OC exposures and implantation failure because it is currently unobservable. The sec ond limitation is that serum concentrations were available only for females and not their male partners. However, exposures to certain ubiquitous chemicals have been shown to be correlated within couples undergoing IVF (Mahalingaiah et al. 2008). Certain predictors of serum OC concentration were not available for this analysis, such as past breastfeeding history and country of origin. However, most patients were likely to be of American origin; other known predictors of persistent pesti cide exposure were included in our models, including age and smoking. Finally, the serum sample from which OC concentration and lipid content were quanti fied was not a fasting sample. Although both lipid content and serum wet weight of OC increase after a meal, they do not increase pro portionally. Furthermore, the method chosen for lipid correction does affect the final cor rected concentration. The equation used in our study to calculate total lipids was chosen from those suggested by Phillips et al. (1989) and accounts for total cholesterol and triglycerides but does not diminish the differences between fasting and non fasting lipid profiles. However, when lipids are accounted for, there are data to suggest that there still is a high degree of corre lation between a fasting and non fasting serum sample (0.96) (Longnecker et al. 1996).
Our study had a number of strengths. The prospective cohort from which this nested case-control study population was drawn employed risk set sampling that allowed for collection of questionnaire data and biological samples before the cycle outcome and tem porally close to the clinically and biologically relevant outcomes of interest. In addition, our analyses applied advanced statistical methodol ogy that validly accommodated inclusion of data from multiple cycle outcomes per woman and timevarying exposures and potential con founders. The assessment of early reproductive end points observable only during the IVF process allows for insight into critical points in early pregnancy.
The HCB levels in the study population were close to but generally lower than those of the adult female U.S. population (CDC 2009). Additionally, levels of DDT/DDE observed within Q4 among the women in our study population were lower than those of the general populations of many Asian and African countries where DDT is still in use (CDC 2009;Perry et al. 2006;Rusiecki et al. 2005). Therefore, relations observed within our study population may be estimating the association of these OCs only within the lower tail of the international exposure distribution. Further studies in the general U.S. population and in countries with high pesticide burdens are needed.

Conclusion
We found significantly greater odds of implan tation failure in relation to serum HCB levels. Persistent organic chemicals, despite banned use within the United States, may have an adverse impact on conception and pregnancy progres sion. Couplebased exposure assessment and study of the adverse effects on gametes, endo metrium, and embryo development are needed so that these risks can be better understood.