Association between neighborhood poverty and ovarian reserve: The Ovarian Aging Study

Abstract

Objective:

To examine the association between neighborhood poverty and ovarian reserve.

Methods:

Among 1019 healthy, premenopausal women in the Ovarian Aging (OVA) Study, aggregate exposure to neighborhood poverty was examined in relation to biomarkers of ovarian reserve, anti-mullerian hormone (AMH) and antral follicle count (AFC). Specifically, the interaction of age-x-neighborhood poverty was assessed cross-sectionally to determine whether AMH and AFC declines across women may be greater in women exposed to more neighborhood poverty. Neighborhood poverty was assessed by geocoding and linking women’s residential addresses in adulthood to US Census data.

Results:

Independent of covariates, a significant interaction term showed the association between age and AMH varied by degree of exposure to neighborhood poverty in adulthood (b=−0.001, p<.05). AMH declines increased progressively across women exposed to low, medium, and high levels of neighborhood poverty. In addition, main effects showed higher neighborhood poverty was related to higher AMH in the younger women only (b=0.022, p<.01). Results related to AFC were all non-significant (ps>.05).

Conclusions:

Across women, greater aggregate exposure to neighborhood poverty in adulthood was related to lower ovarian reserve, indexed by AMH. In addition, there was a positive association between neighborhood poverty and AMH in younger women that attenuated in the older women. Together, results suggest neighborhood disadvantage may have detrimental impacts that manifest as initially higher AMH, resulting in greater ovarian follicle loss over time. However, it remains unclear whether these results examining differences across women may replicate when AMH declines by neighborhood poverty are examined longitudinally.

Age at the onset of menopause has important health implications (1, 2). Earlier menopausal onset is associated with a variety of health outcomes including osteoporosis, neurological disorders, diabetes, cardiovascular diseases, and early mortality (3, 4). Evidence suggests the exhaustion of the ovarian follicle pool leads to menopause (5, 6). Therefore, a lower ovarian reserve may confer risk for an earlier onset of menopause and associated health risks (7).

Prior studies show women with lower individual-level socioeconomic status (SES) have lower ovarian reserve and an earlier age at menopause (8-18). For example, in the Study of Women’s Health Across the Nation (SWAN), a large multiethnic sample of women, lower educational attainment and unemployment status were associated with earlier menopausal onset independent of race/ethnicity, smoking, use of oral contraceptives, parity, marital status, and history of heart disease (18). In another study examining SES over the life course, women who were the most disadvantaged, defined as having 9 or 10 indicators of low SES over periods of childhood and adulthood, experienced menopause 1.7 years earlier than women who were the least disadvantaged, defined as having 0 or 1 indicator of low SES (9). Other studies have considered how urban vs. rural environments may impact ovarian aging and age at menopause, although results have been inconsistent (19, 20).

In parallel, growing evidence shows neighborhood-level disadvantage has detrimental impacts on women’s health more generally, possibly through health-related behaviors, resource availability, and environmental toxicants (21-24). These same exposures have implications for women’s reproductive health (25), yet few studies have explored links between neighborhood-level SES and ovarian reserve specifically. One prior study showed lower neighborhood-level SES among mothers was associated with lower ovarian reserve in their adult daughters (26), indicating SES-related exposures in utero may influence the size or health of offspring ovarian reserve. One other more recent study showed that among obese or overweight women, those living in the most disadvantaged neighborhoods had a lower ovarian reserve compared to those living in less disadvantaged neighborhoods (27).

Estimating ovarian reserve in women poses numerous challenges (28). However, methodological advances have enabled the measurement of the real-time loss of ovarian follicles underlying variability in the timing of menopause. Using such methods, the number of ovarian follicles remaining in the primordial pool (or ovarian reserve) is able to be estimated. Two biomarkers of ovarian reserve have emerged. AMH, a biochemical marker, is secreted by the granulosa cells of the pre-antral and small antral follicles of the ovary and reflects the number of small growing follicles, which are proportional to the number of primordial follicles remaining in the ovary (29-34). AFC, an ultrasound-derived marker, refers to the number of follicles between 2 and 10 mm that are visible by transvaginal ultrasound in the early follicular phase and are proportional to the number of primordial follicles remaining in the ovary (35-37).

The current study builds on the literatures described above by focusing uniquely on socioeconomic disadvantage at the neighborhood level in relation to estimates of ovarian reserve. Specifically, aggregate exposure to neighborhood poverty over the period of adulthood was examined cross-sectionally in relation to ovarian reserve, indexed by biomarkers AMH and AFC, in a large (N=1019), multiethnic sample of healthy, premenopausal women in the Ovarian Aging (OVA) Study. This work represents a significant, albeit incremental, step forward in addressing a prominent gap in our current knowledge regarding whether neighborhood disadvantage may impact the ovarian reserve and the timing of menopause, which has important implications for women’s health and well-being broadly and may point to areas for potential intervention.

Materials and methods:

Participants:

The current sample included participants in the OVA Study, a community-based, cross-sectional investigation of ovarian aging that entailed a one-time, in-person research visit between 2006 and 2011. The OVA study recruited women who were members of Kaiser Permanente (KP) in Northern California. KP provides healthcare to more than 30% of the population in Northern California (38) and the sociodemographic and health characteristics of the members of KP are generally representative of the Northern California population (39).

Selection criteria for the OVA study were ages between 25 and 45 years, self-identification as one of five race/ethnicities (white, black, Latina, Chinese, and Filipina), and ability to speak/read English, Spanish, or Cantonese. In addition, all participants were required to have regular menstrual cycles and their uterus and both ovaries intact. Exclusion criteria were self-report of major medical illnesses (cardiovascular diseases, chronic kidney or liver disease, diabetes, invasive cancer, chemotherapy or radiation therapy, epilepsy, systemic lupus erythematosus, or HIV-positive status), taking medications impacting the menstrual cycle within 3 months prior to study participation, and current pregnancy or breastfeeding. In brief, the OVA study protocol entailed one in-person research visit to the University of California San Francisco campus, located in San Francisco, California, which included an interview, anthropometric assessment, blood draw, and transvaginal ultrasound (TVUS). The research visits were led by trained staff bilingual in English/Spanish or English/Cantonese and the TVUS exams were all performed by one of two trained reproductive endocrinologists.

A total of 1019 women completed the OVA study protocol. One woman was missing valid residential address information, leaving 1018 women available for analysis in the current study. In addition, in analyses examining AMH, 31 women were not retained (n=987), 26 due to missing AMH data and 5 due to missing information for a covariate of interest. In analyses examining AFC, 46 women were not retained (n=972), 41 due to missing AFC data and 5 due to missing information for a covariate of interest. AMH data were missing due to participant refusal or inadequate blood draw. AFC data were missing due to the presence of an ovarian cyst >30 mm or other masses/anatomical abnormalities obstructing visualization or technical errors.

Institutional review board approval was obtained from KP of Northern California, the University of California San Francisco, and the University of Washington. Written informed consent was obtained from all study participants.

Measures:

Ovarian aging measurement:

AMH:

Blood was collected from each participant between day 2 and 4 of the menstrual cycle. Samples were frozen and shipped in batches to the Central Ligand Assay Satellite Services (CLASS) Laboratory at the University of Michigan for analysis. AMH was assayed using Beckman Coulter’s two-site sandwich enzyme-linked immunosorbent assay (ELISA) (Beckman Coulter, Inc., Brea, CA). Most of the samples (85%) were assayed using this assay. The remaining samples were assayed using the second-generation assay (Gen II) (Beckman Coulter, Inc., Brea, CA). For a subset of 44 participants in whom both assays were performed, the assays showed excellent correspondence (R² = 0.94). AMH values based on the Immunotech assay were adjusted using the equation of the line with Immunotech predicting Gen II. Gen II assay sensitivity was 0.16 ng/ml and the intra- and inter-assay coefficients of variation (CV) were 1.4% and 12.5%, respectively.

AFC:

TVUS was performed on each participant between day 2 and 4 of the menstrual cycle by one of two trained reproductive endocrinologists. A Shimadzu SDU-450XL machine (Shimadzu Scientific Instruments, Columbia, MD) with a variable 4- to 8-mHz vaginal transducer was used to measure the transverse, longitudinal, and anteroposterior diameters of each ovary. Follicles with a mean diameter across two dimensions of 2-10 mm were counted. Each measurement was taken twice and then averaged. The total number of follicles across both ovaries was summed to calculate AFC. Evaluation of a sub-sample of 50 participants showed inter-rater reliability between the two reproductive endocrinologists was excellent (r = 0.92) as was test-retest reliability for each reproductive endocrinologist measured over 2 consecutive months (average r = 0.91).

Neighborhood poverty:

Women’s addresses in adulthood (age 18 to current age) were obtained by querying participants about their current residential address and by purchasing historical addresses from the commercial service LexisNexis. LexisNexis provided the street address, state, zip code, and the beginning and end dates of residence for each available address. LexisNexis was used due to advantages highlighted in a prior report, including the provision of a high level of data security, low expense, and accurate information (40). The participants’ addresses were then geocoded and linked to US Census data. Only one residential address per year was used. In cases of multiple addresses in the same year, the first address was selected. For intervening years between the provided residential addresses, location information was filled in using the last geocoded address.

Address-linked neighborhood SES information was extracted using the Neighborhood Change Database (NCDB) from 1970-2010 (41, 42). This database accounts for changes in the geographical boundaries of census tracts over time. The neighborhood poverty variable was selected as an indicator of neighborhood disadvantage. This variable was defined as the percentage of families within the tract who were living below the poverty line with higher values indicating greater neighborhood disadvantage. Aggregate exposure to neighborhood disadvantage was then estimated for each participant by taking the mean of annual neighborhood-level poverty values between ages 18 years and the current age (age at study participation).

Covariates:

Variables modeled as covariates included race/ethnicity, individual-level education and household income, BMI, and smoking status. Categories of race/ethnicity, individual-level education and household income, and smoking status were derived from participant self-reports as a part of a standardized social and medical history interview. BMI was derived from measurements of participant height and weight obtained as a part of standardized anthropometric assessment. See additional details regarding the coding of these variables in the statistical analysis plan below.

Statistical analysis plan:

All analyses were performed using cross-sectional data. The goal was to examine neighborhood poverty in relation to ovarian aging, indexed, in separate models, by biomarkers of ovarian reserve, AMH and AFC. The analytical approach was to fit linear regression models examining the interaction of age by neighborhood poverty in relation to AMH and AFC outcomes, adjusted for covariates. This approach was taken to determine whether the expected decline of AMH and AFC values across women of increasing age may show a differential pattern of lower ovarian reserve or greater ovarian follicle ‘loss’ with greater exposure to neighborhood poverty. In this way, the known decline of ovarian reserve across women of increasing age is being leveraged to determine if such a decline might be greater with greater exposure to environmental adversities such as neighborhood poverty. In the absence of longitudinal data, the examination of AMH and AFC values by neighborhood poverty across women may inform how AMH and AFC values by neighborhood poverty may change within women over time.

Age (in years) and neighborhood poverty (mean of annual neighborhood poverty values) were examined as continuous variables. AMH and AFC were log-transformed to correct right skewness and examined as continuous variables. Models included covariates (race/ethnicity, individual-level education and household income, BMI, and smoking), main effects (age, neighborhood poverty), and the interaction term (age-x-neighborhood poverty). Variables in the interaction term were centered. Centering is a tool used to establish a reference point for each predictor to aid with the meaningful interpretation of results. Centering is performed by subtracting a predetermined constant from all values of the variable, which shifts the scale, but preserves the units. Here, age was centered at age 25 years and neighborhood poverty was centered at the sample mean of neighborhood poverty. This means that the main effects of age on AMH and AFC were examined at the sample mean of neighborhood poverty and the main effects of neighborhood poverty on AMH and AFC were examined at age 25 years.

To elaborate on the age by neighborhood poverty interaction, follow-up analyses were performed to characterize declines in AMH and AFC values across women of increasing age (slopes) divided into categories (tertiles) of low, medium, and high exposure to neighborhood poverty. To elaborate on the main effects of neighborhood poverty on AMH and AFC, further analyses were performed using age-centering (as described above) at additional prespecified ages (30, 35, 40, and 45 years). Finally, covariate-adjusted custom contrasts were also performed comparing mean AMH and AFC values across categories (tertiles) of low, medium, and high neighborhood poverty at prespecified ages (25, 30, 35, 40, and 45 years).

The indicated covariates were coded according to the following. Race/ethnicity categories (white, black, Latina, Chinese, and Filipina) were dummy coded (0 vs. 1) into four variables using white as the reference group. Individual-level SES indicators included education, coded 1=<HS/some HS; 2=HS grad/GED; 3=some college/AA/vocational school; 4=college graduate; 5=graduate school (PhD, MS); 6=professional degree (MD, JD, DDS, MBA) and household income, coded 1=<$5,000; 2=$5,000-$15,999; 3=$16,000-$24,999; 4=$25,000-$34,999; 5=$35,000-$49,999; 6=$50,000-$74,999; 7=$75,000-$99,999; 8=$100,000-$149,999; 9=$150,000-$199,999; 10=$200,000+ and divided by the number of dependents in the household. BMI was calculated as weight (kg)/height (m²) and was log-transformed to correct right skewness. Smoking was coded dichotomously (0=never smoking, 1=current/past smoking).

Results:

Sample characteristics:

In Table 1, demographic, health, ovarian reserve, and neighborhood poverty information is presented. Participants were 35.2 (SD=5.5) years of age on average and were racially/ethnically diverse (27.3% white, 24.2% black, 22.6% Latina, 21.9% Chinese, 4.0% Filipina). Individual-level SES indicators showed 42.5% of women did not receive a college degree and 68.3% of women reported an annual household income less than $75,000. Women were overweight on average (mean BMI=27.0 [SD=7.0]) and 23.3% reported current/past smoking. Ovarian aging indicators showed AMH values were 3.2 (SD=2.8) ng/mL on average and AFC values were 15.1 (SD=9.6) on average. Aggregate exposure to neighborhood poverty showed the percentage of families living below the poverty line was 17.5 (SD=9.7) on average.

Table 1.

Descriptive statistics examining demographic, general and reproductive health, ovarian reserve, and neighborhood poverty information for the full sample (N=1018).

	Mean (SD) or %
Age (y)	35.2 (5.5)
White	27.3%
Black	24.2%
Latina	22.6%
Chinese	21.9%
Filipina	4.0%
Education (% < college grad)	42.5%
Household income (% < $75,000)	68.3%
BMI (kg/m2)	27.0 (7.0)
Smoking (% current/past)	23.3%
Parity (% ≥ 1 live birth)	42.8%
Hormonal contraception (% past use)	70.0%
Menarcheal age (y)	12.6 (1.6)
AMH (ng/mL)	3.2 (2.8)
AFC	15.1 (9.6)
Mean % of families below poverty line	17.5 (9.7)

Missing data: Of the total sample (n=1018), 26 participants were missing AMH data, 41 participants were missing AFC data, and 5 participants were missing household income data. AFC, antral follicle count; AMH, anti-mullerian hormone.

Bivariate correlations:

Bivariate correlations between neighborhood poverty and ovarian aging biomarkers showed greater aggregate exposure to neighborhood poverty was related to lower AMH (r=−0.260; p<0.0001) and lower AFC (r=−0.301; p< 0.0001).

Multivariate analyses:

In Table 2, results of covariate-adjusted linear regression analyses estimating the interaction of age and neighborhood poverty in relation to ovarian aging biomarkers are reported.

Table 2.

Linear regression analyses examining age-x-neighborhood-level poverty effects on AMH and AFC adjusted for covariates.

	log AMH			log AFC
	b a	95% CI	p	b a	95% CI	p
Blackb	−0.129	−0.311 to 0.053	n.s.	0.014	−0.116 to 0.145	n.s.
Latinab	−0.320	−0.509 to −0.132	<0.001	−0.009	−0.144 to 0.126	n.s.
Chineseb	−0.202	−0.374 to −0.031	<0.05	−0.080	−0.202 to 0.042	n.s.
Filipinab	−0.310	−0.600 to −0.021	<0.05	−0.115	−0.324 to 0.093	n.s.
Education	0.018	−0.037 to 0.073	n.s.	0.027	−0.013 to 0.066	n.s.
Income	0.001	−0.030 to 0.033	n.s.	0.007	−0.016 to 0.030	n.s.
BMI	−0.500	−0.779 to −0.220	<0.001	0.006	−0.196 to 0.209	n.s.
Smoking	0.048	−0.087 to 0.183	n.s.	0.014	−0.082 to 0.111	n.s.
Agec	−0.074	−0.096 to −0.053	<0.0001	−0.067	−0.082 to −0.052	<0.0001
Neighborhood povertyd	0.022	0.008 to 0.037	<0.01	0.008	−0.003 to 0.018	n.s.
Age-x-neighborhood poverty	−0.001	−0.002 to −0.0003	<0.05	−0.0004	−0.001to 0.0003	n.s.

^aUnstandardized regression coefficient

^bWhite is the reference group

^cThe effect of age on AMH/AFC was estimated at the sample mean of neighborhood poverty

^dThe effect of neighborhood poverty on AMH/AFC was estimated at age 25. Additional results are provided in the text regarding the effect of neighborhood poverty on AMH/AFC at pre-specified ages (i.e., 25, 30, 35, 40, 45).

AFC, antral follicle count; AMH, anti-mullerian hormone.

AMH.

Results showed the age-x-neighborhood poverty interaction term was statistically significant (b=−0.001, p<.05), indicating that the effect of age on AMH varied over levels of aggregate exposure to neighborhood poverty (Table 2). To further characterize this interaction, in follow-up analyses, the distribution of the neighborhood poverty variable was divided into tertiles (low, medium, and high neighborhood poverty) and the unadjusted association between age and AMH was examined in separate regression models within these categories. As depicted in Figure 1, slopes showed that, across women, those who were exposed to low (b=−0.061 AMH/year across women), medium (b=−0.097 AMH/year across women), and high (b=−0.115 AMH/year across women) levels of neighborhood poverty exhibited increasingly lower ovarian reserve, indexed by AMH (ng/mL, log transformed).

Figure 1.

AMH decline across women exposed to low, medium, and high levels of neighborhood poverty. AMH, anti-mullerian hormone

Inspection of the main effects showed older age was associated with lower AMH (b=−0.074, p<0.0001) as estimated at the sample mean of neighborhood poverty (due to centering the neighborhood poverty variable at the sample mean) and higher neighborhood poverty was associated with higher AMH (b=0.022, p<0.01) as estimated at age 25 (due to centering the age variable at age 25) (Table 2). Of note, centering at age 30 (b=0.015, p<0.01), age 35 (b=0.008, ps>.05), age 40 (b=0.001, ps>.05), and age 45 (b=−0.006, ps>.05) showed the association was no longer significant at age 35 and the direction of association between neighborhood poverty and AMH changed from positive to negative with increasing age (Table 3). To further characterize the main effect of neighborhood poverty on AMH in the full age range of the sample, in follow-up analyses, a series of covariate-adjusted custom contrasts were performed comparing mean AMH across tertiles of low, medium, and high neighborhood poverty at prespecified ages: 25, 30, 35, 40, and 45 years. The results corroborated the findings described above based on age-centering. That is, results showed that, among women aged 25, those who were exposed to high (vs. low) levels of neighborhood poverty (b=0.496, p<0.01) and medium (vs. low) levels of neighborhood poverty (b=0.324, p< 0.05) had higher AMH. Among women aged 30, differences were also present between high (vs. low) levels of neighborhood poverty (b=0.278, p<0.05) but attenuated between medium (vs. low) levels of neighborhood poverty (p>.05). In contrast, there were no differences at the older ages, including ages 35, 40, and 45 years (ps>.05).

Table 3.

Linear regression analyses examining age-x-neighborhood-level poverty effects on AMH and AFC adjusted for covariates with age-centered at 25, 30, 35, 40, and 45.

	DV: Ovarian Aging Indexed by log AMH		DV: Ovarian Aging Indexed by log AFC
Age-centered at (years)	ba [95% CI]	P	ba [95% CI]	p
25	0.022 [0.008 to 0.037]	<0.01	0.008 [−0.003 to 0.018]	n.s
30	0.015 [0.005 to 0.026]	<0.01	0.006 [−0.002 to 0.013]	n.s
35	0.008 [−0.0003 to 0.017]	n.s.	0.003 [−0.003 to 0.009]	n.s
40	0.001 [−0.008 to 0.010]	n.s	0.001 [−0.005 to 0.008]	n.s
45	−0.006 [−0.019 to 0.007]	n.s	−0.001 [−0.010 to 0.008]	n.s

^aUnstandardized regression coefficient

AFC, antral follicle count; AMH, anti-mullerian hormone.

Finally, inspection of the covariates showed racial/ethnic identification as Latina, Chinese, or Filipina (vs. white) was associated with lower AMH (ps<.05) and higher BMI was associated with lower AMH (p=−0.500, p<.001) (Table 2).

In sum, results regarding AMH are two-fold, suggesting that 1) across women, those who were exposed to low, medium, and high levels of neighborhood poverty exhibited increasingly lower ovarian reserve, indexed by AMH and that 2) higher neighborhood poverty was related to higher AMH in the youngest women, but this association attenuated by age 35 and became negative (i.e., higher neighborhood poverty was related to lower AMH) in the older women, although not reaching statistical significance.

AFC.

Results showed the age-x-neighborhood poverty interaction term was not statistically significant (p>.05), indicating, in contrast to the AMH findings, that the effect of age on AFC did not vary significantly over levels of aggregate exposure to neighborhood poverty (Table 2). Although not significant, follow-up analyses were still pursued on an exploratory basis to examine the non-significant interaction. In these analyses, as depicted in Figure 2, slopes showed that, across women, those who were exposed to low (b=−0.058 AFC/year across women), medium (b=−0.078 AFC/year across women), and high (b=−0.081 AFC/year across women) levels of neighborhood poverty exhibited increasingly lower ovarian reserve, indexed by AFC (log transformed). This pattern was less pronounced but resembled the AMH results.Inspection of the main effects showed older age was associated with lower AFC (b=−0.067, p=<0.0001) as estimated at the sample mean of neighborhood poverty (due to centering the neighborhood poverty variable at the sample mean) (Table 1). However, in contrast to the AMH findings, neighborhood poverty (at age 25) was not associated with AFC (b=0.008, p>.05), nor were any of the covariates (ps>.05). Of note, centering at age 30 (b=0.006, ps>.05), age 35 (b=0.003, ps>.05), age 40 (b=0.001, ps>.05), and age 45 (b=−0.001, ps>.05) showed a pattern similar to the AMH findings insofar as the direction of association between neighborhood poverty and AFC changed from positive to negative with increasing age (Table 3). Covariate-adjusted custom contrasts performed to compare mean AFC across tertiles of low, medium, and high neighborhood poverty at prespecified ages: 25, 30, 35, 40, and 45 years were all non-significant (ps>.05).

Figure 2.

AFC decline across women exposed to low, medium, and high levels of neighborhood poverty. AFC, antral follicle count.

In sum, results regarding AFC show the age-x-neighborhood poverty interaction was not statistically significant. However, in exploratory analyses, across women, those who were exposed to low, medium, and high levels of neighborhood poverty exhibited increasingly lower ovarian reserve, indexed by AFC, a pattern similar to the AMH findings.

Discussion:

Prior studies suggest that women from lower SES backgrounds have a lower ovarian reserve and an earlier age at menopause (8-18). Few studies have considered whether SES at the neighborhood level may also impact the ovarian reserve (26-27). The current study sought to fill this gap by leveraging a large sample of premenopausal women in the OVA Study to evaluate whether neighborhood SES is related to the ovarian reserve indexed by biomarkers, AMH and AFC. Neighborhood SES was assessed using US Census data to form an aggregate of exposure to neighborhood poverty over the period of adulthood.

The approach of the current study was to leverage cross-sectional data to examine whether, across women, ovarian reserve estimates may vary by level of exposure to neighborhood poverty. To achieve this, the interaction of age by neighborhood poverty in relation to ovarian reserve outcomes was tested. The results showed an intriguing pattern in which, across women of increasing age, the expected decline in AMH was greater in women exposed to greater neighborhood poverty, even independently of covariates, including race/ethnicity, individual-level education and income, BMI, and smoking. When the distribution of neighborhood poverty was examined categorically, women who were exposed to low, medium, and high levels of neighborhood poverty exhibited increasingly lower AMH values across these groups: 0.061 lower AMH (ng/mL, log) per year older in the low neighborhood poverty group vs. 0.097 lower AMH (ng/mL, log) per year older in the medium neighborhood poverty group vs. 0.115 lower AMH (ng/mL, log) per year older in the high neighborhood poverty group. This pattern of results suggests that declines in AMH across women of increasing age differ by exposure to neighborhood poverty in the direction that greater declines occurred with greater aggregate exposure to neighborhood poverty, supporting the hypothesis that neighborhood disadvantage may have a deleterious impact on the ovarian reserve. However, because these results were based on cross-sectional data, it remains unclear whether this pattern of results examining differences across women may replicate when AMH declines by neighborhood poverty are examined within women over time. Moreover, the overall pattern of results was mixed insofar as the significant interaction term was observed in relation to AMH only. All results related to AFC were non-significant; although, in exploratory analyses, AFC declines by neighborhood poverty across women showed a pattern similar to that of AMH. Finally, in the context of a significant interaction term, the examination of main effects showed that higher neighborhood poverty was associated with higher AMH, but only in the youngest women. In fact, this association attenuated by age 35 and became negative in the older women, although not reaching statistical significance. Specifically, tests at pre-specified ages showed there were significant associations between neighborhood poverty and AMH at ages 25 (b=0.022, p<0.01) and 30 (b=0.015, p<0.01), but not at ages 35 (b=0.008, ps>.05), 40 (b=0.001, ps>.05), and 45 (b=−0.006, ps>.05). Taken together, implications of these results regarding the significant interaction term and the main effects are that neighborhood disadvantage may have detrimental impacts that manifest as initially higher AMH, which, at older ages, result in greater ovarian follicle loss and lower relative AMH values. That is, higher initial AMH values, indicating higher fertility (i.e., more follicles moving out of the ovarian reserve) may be adaptive in the short term but reflect an accelerated rate of depletion of the ovarian reserve over time.

The current pattern of results aligns with prior reports from the OVA Study in which greater psychological stress was associated with higher AFC in younger women and a higher rate of AFC decline across women (43, 44). In parallel, that neighborhood poverty was related to both higher AMH in younger women and greater AMH declines across women suggests that adverse exposures such as neighborhood poverty, along with psychological stress, may disrupt the regulation of the ovarian reserve, appearing to potentiate fertility in the short term, but at the cost of hastening depletion of the ovarian reserve over time. Although speculative, this could lead to earlier menopausal onset and associated increased risk for diseases of aging occurring in the post-menopausal period. This observed ‘trade-off’ is consistent with life history theory. Life history theory describes broadly that reproductive aging and the timing of major reproductive events are shaped by selection pressures to reproduce (45, 46). In adverse or resource-scarce environments, investments in reproduction are prioritized even at other costs, accelerating the pace of reproductive events. This framework has been elaborated in relation to understanding variability in the timing of pubertal onset, for example, suggesting early adversity signals earlier reproductive readiness in girls even at significant social and health costs (47-50). Here, this framework may be extended to understanding variability in fertility and the pace of ovarian follicle loss through adult life. Accordingly, it is possible that adversities indexed by neighborhood poverty may enhance fertility in the short term, but at the cost of ‘wasting’ follicles that ultimately leads to the faster depletion of the ovarian reserve, earlier menopausal onset, and associated health risks.

Implications of the current study are that aspects of disadvantaged neighborhoods may ‘get into the body’ to impact the ovarian reserve. However, the potential biological mechanisms explaining these impacts need clarification. Numerous studies suggest there is a general association between neighborhood disadvantage and health (51, 52, 53). Plausibly, this could extend to the health of the ovarian reserve directly or indirectly. More specifically, evidence suggests that toxicant exposures cluster in low SES environments (54, 55, 56). Endocrine-disrupting chemicals (EDCs) are common in personal care and household products and may be particularly elevated among lower SES individuals (57, 58, 59). With respect to the ovarian reserve, examination of human and animal literatures suggests environmental toxicants may accelerate folliculogenesis and follicular atresia, including in the primordial stage and extending across the spectrum of ovarian follicle development (60), with mounting evidence pointing to impacts of specific EDCs such as arsenic (61-64). The impact of EDCs on the acceleration of folliculogenesis and follicular atresia in particular is consistent with the current finding showing an initially higher AMH, which may result in greater ovarian follicle loss over time. Moreover, growing evidence shows traffic pollutants, indexed by PM2.5 as well as shorter distances to roadways and longer distances to green spaces (i.e., ‘buffer’ zones), are associated with estimates of lower ovarian reserve (65-68).

Several additional patterns of results in the current study deserve attention in future investigations. First, in the current multivariate models, individual-level educational attainment and household income were not independently associated with the ovarian reserve biomarkers. This is in contrast to prior studies reporting associations between individual-level SES and ovarian reserve and menopausal timing. More work is needed to delineate the specific SES indicators that may have particular salience, the age of the sample being studied, and the possibility that individual-level SES, as was observed with respect to neighborhood SES, may impact the ovarian reserve differently at younger versus older ages. Next, the neighborhood SES marker examined in the current study reflected a composite of exposure to neighborhood poverty in adulthood. More work is needed to build on this by examining nuances in the nature of such exposures, including the duration of time spent in adverse neighborhood conditions as well as the movement into and out of neighborhoods. Finally, findings in the current study were only significant in relation to AMH. It remains unclear why associations were not observed in relation to AFC. More work is needed to consider how AMH and AFC represent different aspects of the ovarian reserve and how this may inform the pathways by which adverse exposures may differentially impact the ovarian reserve as indexed by AMH versus AFC. For example, because AMH reflects both antral follicles and pre-antral follicles, it is possible that declines in AMH are indexing follicle loss at an earlier stage of follicle grow, presaging declines in AFC that would be observable later as AFC reflects the antral follicles only.

There are several strengths of the current study. First, a novel methodological approach was implemented in which residential address information was ascertained over the entire period of adulthood and used to derive ‘objective’ neighborhood information from the US Census. This enabled examination of the aggregate exposure to neighborhood poverty versus a single point in time. Moreover, the analytical model was well specified including covariates related to individual-level SES and individual-level health status indicators. Another strength was that the ovarian reserve was assessed using established biomarkers, AMH and AFC, which provided a more accurate estimation of the ovarian follicle pool (37, 43) versus markers such as menopausal timing. Finally, the current study included a large, ethnically diverse sample of reproductive age women in the OVA Study. These women were healthy, regularly cycling and not taking hormonal contraceptives, lessening potential confounds.

There are also several weaknesses of the current study. First, the study design was cross-sectional limiting the evaluation of prospective associations. In this way, it is not possible to know whether the current pattern of results observed across women will replicate when associations are examined within women over time. Next, neighborhood poverty is a contextual variable derived from the US Census, and it is not possible to know how this variable is capturing the individual experiences of women. At the individual-level, women may have unique experiences that buffer or isolate them from neighborhood-level impacts. In addition, no information was available related to other neighborhood variables that might have impacts on the ovarian reserve or may offset effects of neighborhood poverty such as social networks within the community or other relevant local resources. (69, 70). There was also a lack of information that would have informed potential biological mechanisms between neighborhood poverty and the ovarian reserve such as environmental toxicants. Another limitation was that the residential address information purchased from a commercial service was not independently verified. However, this approach was deemed superior to querying women about their historical addresses, which, for the older participants, would have spanned nearly 30 years. Finally, the participants were geographically limited to northern California and excluded some ethnicities such as South Asian women.

Conclusions:

In conclusion, findings from the current study point to the potential impacts of neighborhood poverty on the ovarian reserve, which, in turn, has implications for the timing of menopause and health risks more generally. More research is needed to move this area forward by employing longitudinal study designs, expanding on the neighborhood measurement, and assessing potential mechanisms of neighborhood impacts such as relevant environmental toxicants.