- Email: [email protected]
Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of oral contraceptive use
EI-03613; No of Pages 4 Environment International xxx (2017) xxx–xxx
Contents lists available at ScienceDirect
Environment International journal homepage: www.elsevier.com/locate/envint
Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of oral contraceptive use Gerard Ngueta a,b, Matthew P. Longnecker c, Miyoung Yoon d, Christopher D. Ruark d, Harvey J. Clewell 3rd Melvin E. Andersen c,d, Marc-André Verner a,b,⁎
c,d
,
a
Department of Occupational and Environmental Health, Université de Montréal, 2375 chemin de la Cote-Sainte-Catherine, Montreal, QC H3T 1A8, Canada Universite de Montreal Public Health Research Institute (IRSPUM), Université de Montréal, 7101, Parc Ave., Montreal, QC H3N 1X7, Canada Ramboll Environ, 6 Davis Dr, Research Triangle Park, NC 27709, USA d ScitoVation, 6 Davis Dr, Research Triangle Park, NC 27709, USA b c
a r t i c l e
i n f o
Article history: Received 21 October 2016 Received in revised form 27 March 2017 Accepted 28 March 2017 Available online xxxx Keywords: Quantitative bias assessment Perfluoroalkyl substances (PFAS) Endometriosis Physiologically based pharmacokinetic (PBPK) modeling Oral contraceptive use Environmental epidemiology Biomarkers
a b s t r a c t An association between serum levels of perfluoroalkyl substances (PFAS) and endometriosis has recently been reported in an epidemiologic study. Oral contraceptive use to treat dysmenorrhea (pelvic pain associated with endometriosis) could potentially influence this association by reducing menstrual fluid loss, a route of excretion for PFAS. In this study, we aimed to evaluate the influence of differential oral contraceptive use on the association between PFAS and endometriosis. We used a published life-stage physiologically based pharmacokinetic (PBPK) model to simulate plasma levels of perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS) from birth to age at study participation (range 18–44 years). In the simulated population, PFAS level distributions matched those for controls in the epidemiologic study. Prevalence and geometric mean duration (standard deviation [SD]) of oral contraceptive use in the simulated women were based on data from the National Health and Nutrition Examination Survey; among the women with endometriosis the values were, respectively, 29% and 6.8 (3.1) years; among those without endometriosis these values were 18% and 5.3 (2.8) years. In simulations, menstrual fluid loss (ml/cycle) in women taking oral contraceptives was assumed to be 56% of loss in non-users. We evaluated the association between simulated plasma PFAS concentration and endometriosis in the simulated population using logistic regression. Based on the simulations, the association between PFAS levels and endometriosis attributable to differential contraceptive use had an odds ratio (95% CI) of 1.05 (1.02, 1.07) for a loge unit increase in PFOA and 1.03 (1.02, 1.05) for PFOS. In comparison, the epidemiologic study reported odds ratios of 1.62 (0.99, 2.66) for PFOA and 1.25 (0.87, 1.80) for PFOS. Our results suggest that the influence of oral contraceptive use on the association between PFAS levels and endometriosis is relatively small. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction A recent epidemiologic study reported a positive and statistically significant association between serum levels of certain perfluoroalkyl substances (PFAS) and endometriosis (Louis et al., 2012). After adjustment for women's age and body mass index, the odds ratio (OR) for the diagnosis of endometriosis in a sample of women seeking clinical care was 1.89 (95% CI: 1.17, 3.06) per loge unit increase in serum
⁎ Corresponding author at: Department of Occupational and Environmental Health, Université de Montréal, 2375 ch. de la Cote-Sainte-Catherine, Suite 4105, Montreal, QC H3T 1A8, Canada. E-mail addresses: [email protected] (G. Ngueta), [email protected] (M.P. Longnecker), [email protected] (M. Yoon), [email protected] (C.D. Ruark), [email protected] (H.J. Clewell), [email protected] (M.E. Andersen), [email protected] (M.-A. Verner).
perfluorooctanoic acid (PFOA) levels. Further adjustment for parity conditional on gravidity attenuated the OR to 1.62 (95% CI: 0.99, 2.66), indicating the potential importance of controlling for reproductive factors. Whether there are other measured or unmeasured factors underlying this association remains unknown. A factor that could contribute to this association is oral contraceptive use for the treatment of painful menstrual cramps (dysmenorrhea) that often occur in women with endometriosis (Peterson et al., 2013). Other than reducing pain, these treatments can substantially reduce menstrual fluid loss (Fraser and McCarron, 1991; Fraser et al., 2011; Larsson et al., 1992), a route of elimination for PFAS (Wong et al., 2014, 2015). Thus, women taking oral contraceptives would tend to have higher serum PFAS levels. Data from the National Health and Nutrition Examination Survey (NHANES) for years 2003–2004 and 2005–2006 indicate that women with endometriosis are more likely to use oral contraceptives, and they tend to use them for a longer period (see below, and Supplementary material). This raises the possibility that oral contraceptive
http://dx.doi.org/10.1016/j.envint.2017.03.023 0160-4120/© 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: Ngueta, G., et al., Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of ..., Environ Int (2017), http://dx.doi.org/10.1016/j.envint.2017.03.023
2
G. Ngueta et al. / Environment International xxx (2017) xxx–xxx
use contributed to the observed association between serum PFAS levels and endometriosis. In the present study, we aimed to quantitatively assess the potential influence of oral contraceptive use on the association between serum PFAS levels and endometriosis using a life-stage physiologically-based pharmacokinetic (PBPK) model (Ruark et al., 2017). Specifically, we simulated plasma PFOA and perfluorooctane sulfonate (PFOS) levels in a simulated population matching the controls from the Louis et al. (2012) study. 2. Methods 2.1. PBPK model To simulate lifetime exposure to PFAS, we used a life-stage PBPK model of PFOA and PFOS (Ruark et al., 2017) (Fig. 1) based on the Wu et al. (2015) model of exposure during childhood and adolescence. As described in Ruark et al. (2017), the Wu et al. (2015) model was extended to include physiological descriptions for glomerular filtration rate and PFAS resorption in the kidney during later life stages. Specifically, glomerular filtration rate was assumed to decline by 0.62% each year after age 30 (Clewell et al., 2002; Wetzels et al., 2007). PFAS resorption in the kidney was assumed to follow the same decline as the glomerular filtration rate (the same assumption was used for model evaluation in the Ruark et al. (2017) study). The description of menstrual plasma equivalent fluid volume was also modified based on a recent report suggesting that the nonblood menstrual fluid has a similar albumin concentration as the blood plasma fraction and should therefore be included in menstrual volume calculations, especially when evaluating excretion of chemicals that are highly bound to albumin (Verner and Longnecker, 2015). Briefly, the PBPK model simulates exposure to PFOA or PFOS and resulting tissue levels from birth to the time of participation in the epidemiologic study (range: 18–44 years). The daily dose (ng/kg/d), which represented the absorbed dose through all exposure routes, varied as a function of women's age (i.e., age-specific intake) and calendar year (i.e., trend in environmental levels). Distribution in the tissues was driven by blood perfusion and tissue:plasma partition coefficients. Excretion of PFOA and PFOS occurred through two different routes: urinary excretion and menstruation. For renal excretion, free PFAS are first extracted
Fig. 1. Conceptual representation of the PBPK model for PFAS.
from the plasma compartment through glomerular filtration. PFAS in the filtrate are subsequently reabsorbed or transferred to the storage compartment prior to excretion in urine. For excretion through menstruation, PFAS concentration in excreted menstrual fluid was assumed to be equal to women's plasma concentration. Model simulations were carried out using acslX (Aegis Technologies Inc., Huntsville, USA). The accuracy of this PBPK model using the same input parameters for Monte Carlo simulations was previously evaluated against data from NHANES (years 1999–2000, 2003–2004, 2005–2006, 2007–2008, 2009–2010) (Ruark et al., 2017). Distributions of plasma PFOS levels matched the age-specific 5th percentile, 95th percentile and geometric mean of measured PFOS levels. For PFOA, age-specific geometric means of simulated plasma PFOA levels matched measured levels, but the 5th– 95th percentile interval was wider than that observed in measured levels.
2.2. Monte Carlo simulations We performed 200,000 Monte Carlo simulations. The prevalence of endometriosis in the simulated population was 40.2%, mimicking the “operative cohort” that was studied (Louis et al., 2012; Peterson et al., 2013). Sensitive PBPK model parameters were identified through local and global sensitivity analyses in a previous effort (Ruark et al., 2017). For each simulation, values were sampled from probabilistic distributions for sensitive physiological parameters (described in (Ruark et al., 2017)) and exposure (Table S1). We calibrated exposure (daily intakes of PFOA and PFOS) in our Monte Carlo simulations based on the reported distributions of measured serum levels in women without endometriosis from the Louis et al. (2012) study. Specifically, we manually changed the daily intake distribution parameters (i.e., mean, SD) in Monte Carlo simulations of women without endometriosis until we matched the tertile distributions of controls from the Louis et al. (2012) study. These distributions of PFAS daily intake were used for women with and without endometriosis in subsequent Monte Carlo simulations. For each simulated subject, oral contraceptive use was assigned based on data from NHANES years 2003–2004 and 2005–2006, years during which the reproductive health questionnaire included a question about endometriosis. We excluded women b 18 or N 45 years old, who were on chemotherapy, dialysis, who were pregnant or breastfeeding, who did not answer the question on endometriosis, or were missing data on oral contraceptive use, leaving 1450 women. In this dataset, women with endometriosis had a 29% chance of being oral contraceptive users at the time of study participation whereas for controls the percentage was 18%. The duration of oral contraceptive use was sampled from lognormal distributions depending on disease status; the geometric mean (SD) for oral contraceptive use was 6.8 (3.1) years for women with endometriosis and 5.3 (2.8) years for controls. During the use of oral contraceptives, menstrual fluid loss was assumed to be 56% of values without oral contraceptive use (Larsson et al., 1992). To evaluate the maximum potential for bias in the PFAS-endometriosis association, we also performed secondary simulations where i) menstrual fluid loss was set to 0 during oral contraceptive use (scenario 1); ii) all women with endometriosis were current oral contraceptive users (the duration was sampled from the distribution described above) and none of the women without endometriosis were oral contraceptive users (scenario 2); and iii) daily exposure rate was assumed to stay constant on a μg/kg/d basis throughout women's lifetime, i.e., exposure was not age-dependent nor calendar year-dependent (scenario 3). Monte Carlo simulations were also used to perform a global sensitivity analysis of model parameters on simulated PFAS concentrations in the context of this study. Correlation coefficients were calculated to evaluate the strength of association between model parameter values and PFAS levels at the end of simulations (equivalent to the time and age of study participation for the epidemiologic study).
Please cite this article as: Ngueta, G., et al., Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of ..., Environ Int (2017), http://dx.doi.org/10.1016/j.envint.2017.03.023
G. Ngueta et al. / Environment International xxx (2017) xxx–xxx
3
2.3. Statistical analyses To replicate the methods used in the Louis et al. (2012) study, we estimated ORs for endometriosis using logistic regression. Simulated PFOS and PFOA concentrations were loge-transformed (x + 1), the same transformation used by Louis et al. (2012), and then introduced into the regression models with age and body mass index as covariates. We used SAS software (version 9.4; SAS institute Inc., Cary, NC) for all statistical analyses, and statistical significance was assumed for a pvalue b 0.05. 3. Results The distributions of age and PFAS levels in our simulated population closely matched those reported for controls in the epidemiologic study population (see Table 1). In global sensitivity analyses, three parameters stood out as being the most influential on PFOA and PFOS levels: renal resorption, binding to albumin and menstrual fluid loss (Fig. 2). The duration of oral contraceptive use was also correlated with PFAS levels, but many other factors including age were more influential. The differential effect of oral contraceptive use on the simulated tertile distribution of PFAS concentrations between cases and controls was not apparent (see Table 1). However, a small association was observed in logistic regression analyses between simulated PFAS levels and endometriosis (see Table 2). Based on the simulations, the association between PFAS levels and endometriosis attributable to differential contraceptive use had an OR of 1.05 (95% CI: 1.02, 1.07) for PFOA and 1.03 (95% CI: 1.02, 1.05) for PFOS. In comparison, the epidemiologic study reported ORs of 1.62 (95% CI: 0.99, 2.66) for PFOA and 1.25 (95% CI: 0.87, 1.80) for PFOS. Using the reported ORs and those based on the analysis of the simulated data, we estimated the ORs that Louis et al. (2012) would have obtained had they adjusted for both current contraceptive use and duration of oral contraceptive use, which we call bias-adjusted ORs. These were calculated by exponentiating the difference between the loge of the reported OR and the loge of the OR based on the simulated data (Ruark et al., 2017). The bias-adjusted ORs were, for PFOA, 1.54 (0.94, 2.53) and, for PFOS, 1.21 (0.84, 1.75). With effect size approximated by OR-1, our results suggest that differential oral contraceptive use in women with and without endometriosis explain 13% of the observed effect size for PFOA and 16% for PFOS. Assuming no menstrual fluid loss during oral contraceptive in simulations (secondary simulations, scenario 1) increased the ORs, but the estimates were still much lower than the OR from the epidemiologic study (see Table 2). Even in the unlikely event that all women with endometriosis were oral contraceptive users and none of the women without endometriosis were oral contraceptive users (secondary simulations, scenario 2), simulations suggested that oral contraceptive use cannot fully Table 1 Characteristics of observed and simulated populations. Observed subjects were the “Operative Sample” from the Louis et al. (2012) study which included patients with or without endometriosis. Simulated subjects are from the main simulations. Endometriosis Characteristics Age (years) ≤24 25–29 30–34 N35 PFOA (ng/ml) 1st tertile 2nd tertile 3rd tertile PFOS (ng/ml) 1st tertile 2nd tertile 3rd tertile
No endometriosis
Observed (%)
Simulated (%)
Observed (%)
Simulated (%)
14 25 23 37
14 25 23 38
12 20 21 48
12 20 21 47
26 35 39
37 34 29
37 34 29
37 34 29
27 34 40
37 33 30
37 35 28
37 33 30
Fig. 2. Global sensitivity analysis of the PFAS PBPK model. The influence of model parameters on plasma PFAS concentration was evaluated through correlation analyses between model parameter values and loge(x + 1)-transformed PFAS level at the end of the simulation. Analyses were based on 200,000 Monte Carlo simulations.
explain the association between PFOA levels and endometriosis. Assuming constant exposure on a μg/kg/d basis in study subjects (secondary simulations, scenario 3) yielded similar results to those obtained in main simulations. 4. Conclusions In this study, we evaluated whether the differential use of oral contraceptives in women with and without endometriosis could contribute to the association between serum PFAS levels and endometriosis. Given the PBPK model structure and parameters used in our simulation, the results suggested that the contribution of oral contraceptive use to this association is relatively small. Based on our simulations, oral contraceptive use would explain 13% to 16% of the association between PFAS and endometriosis. These results were generated using information on oral contraceptive use in women with and without endometriosis from NHANES years 2003–2006. In this nationally representative population, the percentage of women N18 and b45 years of age using oral contraceptives was 18% among women without endometriosis and 29% among those with endometriosis. The influence of contraceptive use on the association between PFAS and endometriosis may be larger in other populations where the difference in oral contraceptive use between women with and without endometriosis is greater. However, when we performed secondary simulations where all women with endometriosis were current oral contraceptive users and none of the women without endometriosis used oral contraceptives, the ORs still did not fully explain the observed associations for PFOA. This suggests that not adjusting for oral contraceptive use results in a relatively small bias in studies of PFAS and endometriosis. Regardless, to minimize bias in epidemiologic analyses, use and duration of oral contraceptives should be considered for inclusion as covariates, as should other factors expected to affect excretion, such as hysterectomy and detailed aspects of reproductive history. Certain limitations of our study ought to be mentioned. The PBPK model we used to simulate PFAS concentrations has not been directly validated with regards to excretion through menstrual fluids. It is therefore possible that we have over- or under-estimated the influence of oral contraceptive use on plasma PFAS levels. However, analysis of NHANES data showed that oral contraceptive use is positively associated with serum levels of PFOA and PFOS, and this association was stronger than that resulting from simulations (see Supplementary material,
Please cite this article as: Ngueta, G., et al., Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of ..., Environ Int (2017), http://dx.doi.org/10.1016/j.envint.2017.03.023
4
G. Ngueta et al. / Environment International xxx (2017) xxx–xxx
Table 2 Observed vs. model predicted ORs (95% confidence intervals) for the association between loge(x + 1)-transformed PFOA and PFOS concentrations in plasma and odds of endometriosis. Main simulations were based on oral contraceptive use prevalence and duration in women with and without endometriosis from NHANES 2003–2006. Compound
Observed
Main simulations
Secondary simulations (scenario 1)a
Secondary simulations (scenario 2)b
Secondary simulations (scenario 3)c
PFOA PFOS
1.62 (0.99, 2.66) 1.25 (0.87, 1.80)
1.05 (1.02, 1.07) 1.03 (1.02, 1.05)
1.10 (1.08, 1.13) 1.09 (1.07, 1.11)
1.38 (1.35, 1.41) 1.35 (1.33, 1.37)
1.04 (1.01, 1.06) 1.02 (1.01, 1.04)
a
Simulations were performed using parameters from the main simulations, but we assumed no menstrual fluid loss during oral contraceptive use. Simulations were performed using parameters from the main simulations, but we assumed that all women with endometriosis were current oral contraceptive users and that none of the women without endometriosis took oral contraceptives. c Simulations were performed using parameters from the main simulations, but we assumed constant exposure on a μg/kg/day basis throughout subjects' life. b
Table S3). This discrepancy may reflect an influence of oral contraceptives on physiological factors that were not accounted for in our PBPK model simulations. The discrepancy suggests our results may underestimate the effect of adjusting for oral contraceptives. Overall, our study suggested that oral contraceptive use does not contribute substantially to the association between serum PFAS levels and endometriosis. Our study adds to the growing body of literature on quantitative bias analysis and shows how pharmacokinetic models can be used to quantify bias related to physiology and pharmacokinetics in epidemiologic studies (Ruark et al., 2017; Song et al., 2016; Verner et al., 2013; Verner et al., 2015; Wu et al., 2015). Furthermore, it shows how pharmacokinetic modeling can improve risk assessment (Sarigiannis et al., 2016) through enhanced interpretation of epidemiologic data (Burns et al., 2014). Acknowledgments This study was supported by the American Chemistry Council Long Range Research Initiative Program and 3M. Marc-André Verner is recipient of an Emerging Researcher Fellowship from the Université de Montréal Public Health Research Institute (IRSPUM). Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.envint.2017.03.023. References Burns, C.J., Wright, J.M., Pierson, J.B., Bateson, T.F., Burstyn, I., Goldstein, D.A., et al., 2014. Evaluating uncertainty to strengthen epidemiologic data for use in human health risk assessments. Environ. Health Perspect. 122, 1160–1165. Clewell, H.J., Teeguarden, J., McDonald, T., Sarangapani, R., Lawrence, G., Covington, T., et al., 2002. Review and evaluation of the potential impact of age- and gender-specific pharmacokinetic differences on tissue dosimetry. Crit. Rev. Toxicol. 32, 329–389. Fraser, I.S., McCarron, G., 1991. Randomized trial of 2 hormonal and 2 prostaglandininhibiting agents in women with a complaint of menorrhagia. Aust. N. Z. J. Obstet. Gynaecol. 31, 66–70. Fraser, I.S., Romer, T., Parke, S., Zeun, S., Mellinger, U., Machlitt, A., et al., 2011. Effective treatment of heavy and/or prolonged menstrual bleeding with an oral contraceptive
containing estradiol valerate and dienogest: a randomized, double-blind phase iii trial. Hum. Reprod. 26, 2698–2708. Larsson, G., Milsom, I., Lindstedt, G., Rybo, G., 1992. The influence of a low-dose combined oral contraceptive on menstrual blood loss and iron status. Contraception 46, 327–334. Louis, G.M., Peterson, C.M., Chen, Z., Hediger, M.L., Croughan, M.S., Sundaram, R., et al., 2012. Perfluorochemicals and endometriosis: the endo study. Epidemiology 23, 799–805. Peterson, C.M., Johnstone, E.B., Hammoud, A.O., Stanford, J.B., Varner, M.W., Kennedy, A., et al., 2013. Risk factors associated with endometriosis: importance of study population for characterizing disease in the endo study. Am. J. Obstet. Gynecol. 208, 451 e451-411. Ruark, C.D., Song, G., Yoon, M., Verner, M.A., Andersen, M.E., Clewell 3rd, H.J., et al., 2017. Quantitative bias analysis for epidemiological associations of perfluoroalkyl substance serum concentrations and early onset of menopause. Environ. Int. 99, 245–254. Sarigiannis, D.A., Karakitsios, S.P., Handakas, E., Simou, K., Solomou, E., Gotti, A., 2016. Integrated exposure and risk characterization of bisphenol-a in europe. Food Chem. Toxicol. 98, 134–147. Song, G., Peeples, C.R., Yoon, M., Wu, H., Verner, M.A., Andersen, M.E., et al., 2016. Pharmacokinetic bias analysis of the epidemiological associations between serum polybrominated diphenyl ether (bde-47) and timing of menarche. Environ. Res. 150, 541–548. Verner, M.A., Longnecker, M.P., 2015. Comment on “enhanced elimination of perfluorooctanesulfonic acid by menstruating women: evidence from populationbased pharmacokinetic modeling”. Environ. Sci. Technol. 49, 5836–5837. Verner, M.A., McDougall, R., Glynn, A., Andersen, M.E., Clewell 3rd, H.J., Longnecker, M.P., 2013. Is the relationship between prenatal exposure to pcb-153 and decreased birth weight attributable to pharmacokinetics? Environ. Health Perspect. 121, 1219–1224. Verner, M.A., Loccisano, A.E., Morken, N.H., Yoon, M., Wu, H., McDougall, R., et al., 2015. Associations of perfluoroalkyl substances (pfas) with lower birth weight: an evaluation of potential confounding by glomerular filtration rate using a physiologically based pharmacokinetic model (pbpk). Environ. Health Perspect. 123, 1317–1324. Wetzels, J.F., Kiemeney, L.A., Swinkels, D.W., Willems, H.L., den Heijer, M., 2007. Age- and gender-specific reference values of estimated gfr in caucasians: the nijmegen biomedical study. Kidney Int. 72, 632–637. Wong, F., MacLeod, M., Mueller, J.F., Cousins, I.T., 2014. Enhanced elimination of perfluorooctane sulfonic acid by menstruating women: evidence from populationbased pharmacokinetic modeling. Environ. Sci. Technol. 48, 8807–8814. Wong, F., MacLeod, M., Mueller, J.F., Cousins, I.T., 2015. Response to comment on “enhanced elimination of perfluorooctane sulfonic acid by menstruating women: evidence from population-based pharmacokinetic modeling”. Environ. Sci. Technol. 49, 5838–5839. Wu, H., Yoon, M., Verner, M.A., Xue, J., Luo, M., Andersen, M.E., et al., 2015. Can the observed association between serum perfluoroalkyl substances and delayed menarche be explained on the basis of puberty-related changes in physiology and pharmacokinetics? Environ. Int. 82, 61–68.
Please cite this article as: Ngueta, G., et al., Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of ..., Environ Int (2017), http://dx.doi.org/10.1016/j.envint.2017.03.023
Contents lists available at ScienceDirect
Environment International journal homepage: www.elsevier.com/locate/envint
Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of oral contraceptive use Gerard Ngueta a,b, Matthew P. Longnecker c, Miyoung Yoon d, Christopher D. Ruark d, Harvey J. Clewell 3rd Melvin E. Andersen c,d, Marc-André Verner a,b,⁎
c,d
,
a
Department of Occupational and Environmental Health, Université de Montréal, 2375 chemin de la Cote-Sainte-Catherine, Montreal, QC H3T 1A8, Canada Universite de Montreal Public Health Research Institute (IRSPUM), Université de Montréal, 7101, Parc Ave., Montreal, QC H3N 1X7, Canada Ramboll Environ, 6 Davis Dr, Research Triangle Park, NC 27709, USA d ScitoVation, 6 Davis Dr, Research Triangle Park, NC 27709, USA b c
a r t i c l e
i n f o
Article history: Received 21 October 2016 Received in revised form 27 March 2017 Accepted 28 March 2017 Available online xxxx Keywords: Quantitative bias assessment Perfluoroalkyl substances (PFAS) Endometriosis Physiologically based pharmacokinetic (PBPK) modeling Oral contraceptive use Environmental epidemiology Biomarkers
a b s t r a c t An association between serum levels of perfluoroalkyl substances (PFAS) and endometriosis has recently been reported in an epidemiologic study. Oral contraceptive use to treat dysmenorrhea (pelvic pain associated with endometriosis) could potentially influence this association by reducing menstrual fluid loss, a route of excretion for PFAS. In this study, we aimed to evaluate the influence of differential oral contraceptive use on the association between PFAS and endometriosis. We used a published life-stage physiologically based pharmacokinetic (PBPK) model to simulate plasma levels of perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS) from birth to age at study participation (range 18–44 years). In the simulated population, PFAS level distributions matched those for controls in the epidemiologic study. Prevalence and geometric mean duration (standard deviation [SD]) of oral contraceptive use in the simulated women were based on data from the National Health and Nutrition Examination Survey; among the women with endometriosis the values were, respectively, 29% and 6.8 (3.1) years; among those without endometriosis these values were 18% and 5.3 (2.8) years. In simulations, menstrual fluid loss (ml/cycle) in women taking oral contraceptives was assumed to be 56% of loss in non-users. We evaluated the association between simulated plasma PFAS concentration and endometriosis in the simulated population using logistic regression. Based on the simulations, the association between PFAS levels and endometriosis attributable to differential contraceptive use had an odds ratio (95% CI) of 1.05 (1.02, 1.07) for a loge unit increase in PFOA and 1.03 (1.02, 1.05) for PFOS. In comparison, the epidemiologic study reported odds ratios of 1.62 (0.99, 2.66) for PFOA and 1.25 (0.87, 1.80) for PFOS. Our results suggest that the influence of oral contraceptive use on the association between PFAS levels and endometriosis is relatively small. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction A recent epidemiologic study reported a positive and statistically significant association between serum levels of certain perfluoroalkyl substances (PFAS) and endometriosis (Louis et al., 2012). After adjustment for women's age and body mass index, the odds ratio (OR) for the diagnosis of endometriosis in a sample of women seeking clinical care was 1.89 (95% CI: 1.17, 3.06) per loge unit increase in serum
⁎ Corresponding author at: Department of Occupational and Environmental Health, Université de Montréal, 2375 ch. de la Cote-Sainte-Catherine, Suite 4105, Montreal, QC H3T 1A8, Canada. E-mail addresses: [email protected] (G. Ngueta), [email protected] (M.P. Longnecker), [email protected] (M. Yoon), [email protected] (C.D. Ruark), [email protected] (H.J. Clewell), [email protected] (M.E. Andersen), [email protected] (M.-A. Verner).
perfluorooctanoic acid (PFOA) levels. Further adjustment for parity conditional on gravidity attenuated the OR to 1.62 (95% CI: 0.99, 2.66), indicating the potential importance of controlling for reproductive factors. Whether there are other measured or unmeasured factors underlying this association remains unknown. A factor that could contribute to this association is oral contraceptive use for the treatment of painful menstrual cramps (dysmenorrhea) that often occur in women with endometriosis (Peterson et al., 2013). Other than reducing pain, these treatments can substantially reduce menstrual fluid loss (Fraser and McCarron, 1991; Fraser et al., 2011; Larsson et al., 1992), a route of elimination for PFAS (Wong et al., 2014, 2015). Thus, women taking oral contraceptives would tend to have higher serum PFAS levels. Data from the National Health and Nutrition Examination Survey (NHANES) for years 2003–2004 and 2005–2006 indicate that women with endometriosis are more likely to use oral contraceptives, and they tend to use them for a longer period (see below, and Supplementary material). This raises the possibility that oral contraceptive
http://dx.doi.org/10.1016/j.envint.2017.03.023 0160-4120/© 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: Ngueta, G., et al., Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of ..., Environ Int (2017), http://dx.doi.org/10.1016/j.envint.2017.03.023
2
G. Ngueta et al. / Environment International xxx (2017) xxx–xxx
use contributed to the observed association between serum PFAS levels and endometriosis. In the present study, we aimed to quantitatively assess the potential influence of oral contraceptive use on the association between serum PFAS levels and endometriosis using a life-stage physiologically-based pharmacokinetic (PBPK) model (Ruark et al., 2017). Specifically, we simulated plasma PFOA and perfluorooctane sulfonate (PFOS) levels in a simulated population matching the controls from the Louis et al. (2012) study. 2. Methods 2.1. PBPK model To simulate lifetime exposure to PFAS, we used a life-stage PBPK model of PFOA and PFOS (Ruark et al., 2017) (Fig. 1) based on the Wu et al. (2015) model of exposure during childhood and adolescence. As described in Ruark et al. (2017), the Wu et al. (2015) model was extended to include physiological descriptions for glomerular filtration rate and PFAS resorption in the kidney during later life stages. Specifically, glomerular filtration rate was assumed to decline by 0.62% each year after age 30 (Clewell et al., 2002; Wetzels et al., 2007). PFAS resorption in the kidney was assumed to follow the same decline as the glomerular filtration rate (the same assumption was used for model evaluation in the Ruark et al. (2017) study). The description of menstrual plasma equivalent fluid volume was also modified based on a recent report suggesting that the nonblood menstrual fluid has a similar albumin concentration as the blood plasma fraction and should therefore be included in menstrual volume calculations, especially when evaluating excretion of chemicals that are highly bound to albumin (Verner and Longnecker, 2015). Briefly, the PBPK model simulates exposure to PFOA or PFOS and resulting tissue levels from birth to the time of participation in the epidemiologic study (range: 18–44 years). The daily dose (ng/kg/d), which represented the absorbed dose through all exposure routes, varied as a function of women's age (i.e., age-specific intake) and calendar year (i.e., trend in environmental levels). Distribution in the tissues was driven by blood perfusion and tissue:plasma partition coefficients. Excretion of PFOA and PFOS occurred through two different routes: urinary excretion and menstruation. For renal excretion, free PFAS are first extracted
Fig. 1. Conceptual representation of the PBPK model for PFAS.
from the plasma compartment through glomerular filtration. PFAS in the filtrate are subsequently reabsorbed or transferred to the storage compartment prior to excretion in urine. For excretion through menstruation, PFAS concentration in excreted menstrual fluid was assumed to be equal to women's plasma concentration. Model simulations were carried out using acslX (Aegis Technologies Inc., Huntsville, USA). The accuracy of this PBPK model using the same input parameters for Monte Carlo simulations was previously evaluated against data from NHANES (years 1999–2000, 2003–2004, 2005–2006, 2007–2008, 2009–2010) (Ruark et al., 2017). Distributions of plasma PFOS levels matched the age-specific 5th percentile, 95th percentile and geometric mean of measured PFOS levels. For PFOA, age-specific geometric means of simulated plasma PFOA levels matched measured levels, but the 5th– 95th percentile interval was wider than that observed in measured levels.
2.2. Monte Carlo simulations We performed 200,000 Monte Carlo simulations. The prevalence of endometriosis in the simulated population was 40.2%, mimicking the “operative cohort” that was studied (Louis et al., 2012; Peterson et al., 2013). Sensitive PBPK model parameters were identified through local and global sensitivity analyses in a previous effort (Ruark et al., 2017). For each simulation, values were sampled from probabilistic distributions for sensitive physiological parameters (described in (Ruark et al., 2017)) and exposure (Table S1). We calibrated exposure (daily intakes of PFOA and PFOS) in our Monte Carlo simulations based on the reported distributions of measured serum levels in women without endometriosis from the Louis et al. (2012) study. Specifically, we manually changed the daily intake distribution parameters (i.e., mean, SD) in Monte Carlo simulations of women without endometriosis until we matched the tertile distributions of controls from the Louis et al. (2012) study. These distributions of PFAS daily intake were used for women with and without endometriosis in subsequent Monte Carlo simulations. For each simulated subject, oral contraceptive use was assigned based on data from NHANES years 2003–2004 and 2005–2006, years during which the reproductive health questionnaire included a question about endometriosis. We excluded women b 18 or N 45 years old, who were on chemotherapy, dialysis, who were pregnant or breastfeeding, who did not answer the question on endometriosis, or were missing data on oral contraceptive use, leaving 1450 women. In this dataset, women with endometriosis had a 29% chance of being oral contraceptive users at the time of study participation whereas for controls the percentage was 18%. The duration of oral contraceptive use was sampled from lognormal distributions depending on disease status; the geometric mean (SD) for oral contraceptive use was 6.8 (3.1) years for women with endometriosis and 5.3 (2.8) years for controls. During the use of oral contraceptives, menstrual fluid loss was assumed to be 56% of values without oral contraceptive use (Larsson et al., 1992). To evaluate the maximum potential for bias in the PFAS-endometriosis association, we also performed secondary simulations where i) menstrual fluid loss was set to 0 during oral contraceptive use (scenario 1); ii) all women with endometriosis were current oral contraceptive users (the duration was sampled from the distribution described above) and none of the women without endometriosis were oral contraceptive users (scenario 2); and iii) daily exposure rate was assumed to stay constant on a μg/kg/d basis throughout women's lifetime, i.e., exposure was not age-dependent nor calendar year-dependent (scenario 3). Monte Carlo simulations were also used to perform a global sensitivity analysis of model parameters on simulated PFAS concentrations in the context of this study. Correlation coefficients were calculated to evaluate the strength of association between model parameter values and PFAS levels at the end of simulations (equivalent to the time and age of study participation for the epidemiologic study).
Please cite this article as: Ngueta, G., et al., Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of ..., Environ Int (2017), http://dx.doi.org/10.1016/j.envint.2017.03.023
G. Ngueta et al. / Environment International xxx (2017) xxx–xxx
3
2.3. Statistical analyses To replicate the methods used in the Louis et al. (2012) study, we estimated ORs for endometriosis using logistic regression. Simulated PFOS and PFOA concentrations were loge-transformed (x + 1), the same transformation used by Louis et al. (2012), and then introduced into the regression models with age and body mass index as covariates. We used SAS software (version 9.4; SAS institute Inc., Cary, NC) for all statistical analyses, and statistical significance was assumed for a pvalue b 0.05. 3. Results The distributions of age and PFAS levels in our simulated population closely matched those reported for controls in the epidemiologic study population (see Table 1). In global sensitivity analyses, three parameters stood out as being the most influential on PFOA and PFOS levels: renal resorption, binding to albumin and menstrual fluid loss (Fig. 2). The duration of oral contraceptive use was also correlated with PFAS levels, but many other factors including age were more influential. The differential effect of oral contraceptive use on the simulated tertile distribution of PFAS concentrations between cases and controls was not apparent (see Table 1). However, a small association was observed in logistic regression analyses between simulated PFAS levels and endometriosis (see Table 2). Based on the simulations, the association between PFAS levels and endometriosis attributable to differential contraceptive use had an OR of 1.05 (95% CI: 1.02, 1.07) for PFOA and 1.03 (95% CI: 1.02, 1.05) for PFOS. In comparison, the epidemiologic study reported ORs of 1.62 (95% CI: 0.99, 2.66) for PFOA and 1.25 (95% CI: 0.87, 1.80) for PFOS. Using the reported ORs and those based on the analysis of the simulated data, we estimated the ORs that Louis et al. (2012) would have obtained had they adjusted for both current contraceptive use and duration of oral contraceptive use, which we call bias-adjusted ORs. These were calculated by exponentiating the difference between the loge of the reported OR and the loge of the OR based on the simulated data (Ruark et al., 2017). The bias-adjusted ORs were, for PFOA, 1.54 (0.94, 2.53) and, for PFOS, 1.21 (0.84, 1.75). With effect size approximated by OR-1, our results suggest that differential oral contraceptive use in women with and without endometriosis explain 13% of the observed effect size for PFOA and 16% for PFOS. Assuming no menstrual fluid loss during oral contraceptive in simulations (secondary simulations, scenario 1) increased the ORs, but the estimates were still much lower than the OR from the epidemiologic study (see Table 2). Even in the unlikely event that all women with endometriosis were oral contraceptive users and none of the women without endometriosis were oral contraceptive users (secondary simulations, scenario 2), simulations suggested that oral contraceptive use cannot fully Table 1 Characteristics of observed and simulated populations. Observed subjects were the “Operative Sample” from the Louis et al. (2012) study which included patients with or without endometriosis. Simulated subjects are from the main simulations. Endometriosis Characteristics Age (years) ≤24 25–29 30–34 N35 PFOA (ng/ml) 1st tertile 2nd tertile 3rd tertile PFOS (ng/ml) 1st tertile 2nd tertile 3rd tertile
No endometriosis
Observed (%)
Simulated (%)
Observed (%)
Simulated (%)
14 25 23 37
14 25 23 38
12 20 21 48
12 20 21 47
26 35 39
37 34 29
37 34 29
37 34 29
27 34 40
37 33 30
37 35 28
37 33 30
Fig. 2. Global sensitivity analysis of the PFAS PBPK model. The influence of model parameters on plasma PFAS concentration was evaluated through correlation analyses between model parameter values and loge(x + 1)-transformed PFAS level at the end of the simulation. Analyses were based on 200,000 Monte Carlo simulations.
explain the association between PFOA levels and endometriosis. Assuming constant exposure on a μg/kg/d basis in study subjects (secondary simulations, scenario 3) yielded similar results to those obtained in main simulations. 4. Conclusions In this study, we evaluated whether the differential use of oral contraceptives in women with and without endometriosis could contribute to the association between serum PFAS levels and endometriosis. Given the PBPK model structure and parameters used in our simulation, the results suggested that the contribution of oral contraceptive use to this association is relatively small. Based on our simulations, oral contraceptive use would explain 13% to 16% of the association between PFAS and endometriosis. These results were generated using information on oral contraceptive use in women with and without endometriosis from NHANES years 2003–2006. In this nationally representative population, the percentage of women N18 and b45 years of age using oral contraceptives was 18% among women without endometriosis and 29% among those with endometriosis. The influence of contraceptive use on the association between PFAS and endometriosis may be larger in other populations where the difference in oral contraceptive use between women with and without endometriosis is greater. However, when we performed secondary simulations where all women with endometriosis were current oral contraceptive users and none of the women without endometriosis used oral contraceptives, the ORs still did not fully explain the observed associations for PFOA. This suggests that not adjusting for oral contraceptive use results in a relatively small bias in studies of PFAS and endometriosis. Regardless, to minimize bias in epidemiologic analyses, use and duration of oral contraceptives should be considered for inclusion as covariates, as should other factors expected to affect excretion, such as hysterectomy and detailed aspects of reproductive history. Certain limitations of our study ought to be mentioned. The PBPK model we used to simulate PFAS concentrations has not been directly validated with regards to excretion through menstrual fluids. It is therefore possible that we have over- or under-estimated the influence of oral contraceptive use on plasma PFAS levels. However, analysis of NHANES data showed that oral contraceptive use is positively associated with serum levels of PFOA and PFOS, and this association was stronger than that resulting from simulations (see Supplementary material,
Please cite this article as: Ngueta, G., et al., Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of ..., Environ Int (2017), http://dx.doi.org/10.1016/j.envint.2017.03.023
4
G. Ngueta et al. / Environment International xxx (2017) xxx–xxx
Table 2 Observed vs. model predicted ORs (95% confidence intervals) for the association between loge(x + 1)-transformed PFOA and PFOS concentrations in plasma and odds of endometriosis. Main simulations were based on oral contraceptive use prevalence and duration in women with and without endometriosis from NHANES 2003–2006. Compound
Observed
Main simulations
Secondary simulations (scenario 1)a
Secondary simulations (scenario 2)b
Secondary simulations (scenario 3)c
PFOA PFOS
1.62 (0.99, 2.66) 1.25 (0.87, 1.80)
1.05 (1.02, 1.07) 1.03 (1.02, 1.05)
1.10 (1.08, 1.13) 1.09 (1.07, 1.11)
1.38 (1.35, 1.41) 1.35 (1.33, 1.37)
1.04 (1.01, 1.06) 1.02 (1.01, 1.04)
a
Simulations were performed using parameters from the main simulations, but we assumed no menstrual fluid loss during oral contraceptive use. Simulations were performed using parameters from the main simulations, but we assumed that all women with endometriosis were current oral contraceptive users and that none of the women without endometriosis took oral contraceptives. c Simulations were performed using parameters from the main simulations, but we assumed constant exposure on a μg/kg/day basis throughout subjects' life. b
Table S3). This discrepancy may reflect an influence of oral contraceptives on physiological factors that were not accounted for in our PBPK model simulations. The discrepancy suggests our results may underestimate the effect of adjusting for oral contraceptives. Overall, our study suggested that oral contraceptive use does not contribute substantially to the association between serum PFAS levels and endometriosis. Our study adds to the growing body of literature on quantitative bias analysis and shows how pharmacokinetic models can be used to quantify bias related to physiology and pharmacokinetics in epidemiologic studies (Ruark et al., 2017; Song et al., 2016; Verner et al., 2013; Verner et al., 2015; Wu et al., 2015). Furthermore, it shows how pharmacokinetic modeling can improve risk assessment (Sarigiannis et al., 2016) through enhanced interpretation of epidemiologic data (Burns et al., 2014). Acknowledgments This study was supported by the American Chemistry Council Long Range Research Initiative Program and 3M. Marc-André Verner is recipient of an Emerging Researcher Fellowship from the Université de Montréal Public Health Research Institute (IRSPUM). Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.envint.2017.03.023. References Burns, C.J., Wright, J.M., Pierson, J.B., Bateson, T.F., Burstyn, I., Goldstein, D.A., et al., 2014. Evaluating uncertainty to strengthen epidemiologic data for use in human health risk assessments. Environ. Health Perspect. 122, 1160–1165. Clewell, H.J., Teeguarden, J., McDonald, T., Sarangapani, R., Lawrence, G., Covington, T., et al., 2002. Review and evaluation of the potential impact of age- and gender-specific pharmacokinetic differences on tissue dosimetry. Crit. Rev. Toxicol. 32, 329–389. Fraser, I.S., McCarron, G., 1991. Randomized trial of 2 hormonal and 2 prostaglandininhibiting agents in women with a complaint of menorrhagia. Aust. N. Z. J. Obstet. Gynaecol. 31, 66–70. Fraser, I.S., Romer, T., Parke, S., Zeun, S., Mellinger, U., Machlitt, A., et al., 2011. Effective treatment of heavy and/or prolonged menstrual bleeding with an oral contraceptive
containing estradiol valerate and dienogest: a randomized, double-blind phase iii trial. Hum. Reprod. 26, 2698–2708. Larsson, G., Milsom, I., Lindstedt, G., Rybo, G., 1992. The influence of a low-dose combined oral contraceptive on menstrual blood loss and iron status. Contraception 46, 327–334. Louis, G.M., Peterson, C.M., Chen, Z., Hediger, M.L., Croughan, M.S., Sundaram, R., et al., 2012. Perfluorochemicals and endometriosis: the endo study. Epidemiology 23, 799–805. Peterson, C.M., Johnstone, E.B., Hammoud, A.O., Stanford, J.B., Varner, M.W., Kennedy, A., et al., 2013. Risk factors associated with endometriosis: importance of study population for characterizing disease in the endo study. Am. J. Obstet. Gynecol. 208, 451 e451-411. Ruark, C.D., Song, G., Yoon, M., Verner, M.A., Andersen, M.E., Clewell 3rd, H.J., et al., 2017. Quantitative bias analysis for epidemiological associations of perfluoroalkyl substance serum concentrations and early onset of menopause. Environ. Int. 99, 245–254. Sarigiannis, D.A., Karakitsios, S.P., Handakas, E., Simou, K., Solomou, E., Gotti, A., 2016. Integrated exposure and risk characterization of bisphenol-a in europe. Food Chem. Toxicol. 98, 134–147. Song, G., Peeples, C.R., Yoon, M., Wu, H., Verner, M.A., Andersen, M.E., et al., 2016. Pharmacokinetic bias analysis of the epidemiological associations between serum polybrominated diphenyl ether (bde-47) and timing of menarche. Environ. Res. 150, 541–548. Verner, M.A., Longnecker, M.P., 2015. Comment on “enhanced elimination of perfluorooctanesulfonic acid by menstruating women: evidence from populationbased pharmacokinetic modeling”. Environ. Sci. Technol. 49, 5836–5837. Verner, M.A., McDougall, R., Glynn, A., Andersen, M.E., Clewell 3rd, H.J., Longnecker, M.P., 2013. Is the relationship between prenatal exposure to pcb-153 and decreased birth weight attributable to pharmacokinetics? Environ. Health Perspect. 121, 1219–1224. Verner, M.A., Loccisano, A.E., Morken, N.H., Yoon, M., Wu, H., McDougall, R., et al., 2015. Associations of perfluoroalkyl substances (pfas) with lower birth weight: an evaluation of potential confounding by glomerular filtration rate using a physiologically based pharmacokinetic model (pbpk). Environ. Health Perspect. 123, 1317–1324. Wetzels, J.F., Kiemeney, L.A., Swinkels, D.W., Willems, H.L., den Heijer, M., 2007. Age- and gender-specific reference values of estimated gfr in caucasians: the nijmegen biomedical study. Kidney Int. 72, 632–637. Wong, F., MacLeod, M., Mueller, J.F., Cousins, I.T., 2014. Enhanced elimination of perfluorooctane sulfonic acid by menstruating women: evidence from populationbased pharmacokinetic modeling. Environ. Sci. Technol. 48, 8807–8814. Wong, F., MacLeod, M., Mueller, J.F., Cousins, I.T., 2015. Response to comment on “enhanced elimination of perfluorooctane sulfonic acid by menstruating women: evidence from population-based pharmacokinetic modeling”. Environ. Sci. Technol. 49, 5838–5839. Wu, H., Yoon, M., Verner, M.A., Xue, J., Luo, M., Andersen, M.E., et al., 2015. Can the observed association between serum perfluoroalkyl substances and delayed menarche be explained on the basis of puberty-related changes in physiology and pharmacokinetics? Environ. Int. 82, 61–68.
Please cite this article as: Ngueta, G., et al., Quantitative bias analysis of a reported association between perfluoroalkyl substances (PFAS) and endometriosis: The influence of ..., Environ Int (2017), http://dx.doi.org/10.1016/j.envint.2017.03.023