The Weathering Hypothesis As an Explanation for Racial Disparities in Health: a Systematic Review

Annals of Epidemiology 33 (2019) 1e18

Contents lists available at ScienceDirect

Annals of Epidemiology

Review article The weathering hypothesis as an explanation for racial disparities in health: a systematic review

* Allana T. Forde, PhD, MPH a, , Danielle M. Crookes, MPH a, Shakira F. Suglia, ScD a, b, Ryan T. Demmer, PhD a, c a Department of Epidemiology, Columbia University Mailman School of Public Health, New York, NY b Department of Epidemiology, Emory University Rollins School of Public Health, Atlanta, GA c Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, Minneapolis article info abstract

Article history: Purpose: The weathering hypothesis states that chronic exposure to social and economic disadvantage Received 1 February 2018 leads to accelerated decline in physical health outcomes and could partially explain racial disparities in a Accepted 28 February 2019 wide array of health conditions. This systematic review summarizes the literature empirically testing the Available online 19 March 2019 weathering hypothesis and assesses the quality of the evidence regarding weathering as a determinant of racial disparities in health. Keywords: Methods: Databases (Web of Science, Ovid MEDLINE, PubMed, and Embase) were searched for studies Weathering published in English up to July 1, 2017. Studies that tested the weathering hypothesis for any physical Race Health disparities health outcome and included at least one socially or economically disadvantaged group (e.g., Blacks) for Health inequalities whom the weathering hypothesis applies were assessed for eligibility. Threats to validity were assessed using the Quality in Prognostic Studies tool. Results: The 41 included studies were rated as having overall good methodological quality. Most studies found evidence in support of the weathering hypothesis, although the magnitude of support varied by the health outcome and population studied. Conclusions: Future evaluations of the weathering hypothesis should include an examination of additional health outcomes and interrogate mechanisms that could link weathering to poor health. © 2019 Elsevier Inc. All rights reserved.

Introduction birth outcomes in Blacks relative to Whites [16] and originated from Geronimus' empirical studies on racial disparities in birth Blacks have higher rates of morbidity and mortality than Whites outcomes [15]. Contrary to the well-documented curvilinear rela- for almost all health outcomes in the United States, and this tionship between maternal age and birth outcomes, where teenage inequality increases with age [1e8]. Although these racial dispar- and 30-year-old pregnant women are expected to have a higher ities are notable, their underlying causes are unclear [9,10]. The risk for adverse birth outcomes than women in their mid-20s, weathering hypothesis was proposed to explain racial health dis- studies by Geronimus revealed variations by race. She found that parities. Weathering is the result of chronic exposure to social and White teenage mothers had a higher risk for infant mortality and economic disadvantage that leads to the acceleration of normal low birth weight than White mothers in their mid-to-late 20s, but aging and earlier onset of unfavorable physical health conditions Black teenage mothers had a lower risk for infant mortality and low among disadvantaged (vs. advantaged) persons of similar age (i.e., birth weight than older Black mothers [14,15,17]. weathering pattern) [11e15]. Many studies use this hypothesis as an explanatory framework for The weathering hypothesis was motivated by observations of racial disparities in health, but a systematic review of the studies that earlier onset of chronic diseases (e.g., hypertension) impacting explicitly test this hypothesis has not been previously conducted. The goals of this systematic review are to (1) provide an overview of the existing literature that empirically tests the weathering hypothesis across a variety of physical health outcomes; (2) assess the evidence * Corresponding author. Department of Epidemiology, Columbia University Mailman School of Public Health, 722 West 168th Street, New York, NY 10032. for weathering as a determinant of racial disparities in health; and (3) E-mail address: [email protected] (A.T. Forde). evaluate the threats to validity of existing studies. https://doi.org/10.1016/j.annepidem.2019.02.011 1047-2797/© 2019 Elsevier Inc. All rights reserved. 2 A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18

Methods evaluating the likelihood of bias. When reviewers disagreed on bias assessments, the studies in question (n ¼ 4) were formally Search strategy discussed by both reviewers to arrive at a ﬁnal decision.

The systematic review methodology and reporting followed the Results Preferred Reporting Items for Systematic Reviews and Meta- Analysis guidelines for a review that does not contain a meta- The citation and database searches identified 817 records analysis of studies [18]. First, a citation search was conducted in (Fig. 1). After removal of duplicates, 617 articles were eligible for the Web of Science to identify all articles published by Geronimus and title and abstract review, of which, 341 articles underwent full-text articles that cited her original publications [14,15] describing the review. Of these, 298 were excluded because they did not test the weathering hypothesis. Second, keyword and medical subject weathering hypothesis. Three studies were excluded because they headings searches in Ovid MEDLINE (MeSH), PubMed (MeSH), and lacked a comparison group. One additional article was added after Embase (Emtree) electronic databases were conducted from the reviewing the reference lists. In total, 41 articles were included in first mention of the hypothesis in 1986 to July 1, 2017; the keyword this review. “weathering” was combined with the keyword “racial disparities” and MeSH terms (e.g., health status disparities, health care disparities, inequalities, ethnic group, race difference) to identify Study and participant characteristics relevant articles for the systematic review. Third, the reference lists of the (1) included studies, (2) conceptual/review articles identified The study and participant characteristics are presented in in the search process, and (3) grey literature (nonepeer reviewed) Table 1. The 41 studies in the review were published between 1996 were examined to identify additional studies. All retrieved articles and 2017. Most studies (78%) used a cross-sectional design and the were imported and stored in an EndNote database. remaining 22% of studies used a cohort design. Only two studies were conducted outside the United States (Taiwan and the United Inclusion and exclusion criteria Kingdom). Study participant data were obtained from vital statistics and/or census data alone in 46% of studies or solely from survey Inclusion criteria for articles included in this review were as data in 54% of studies. Sample sizes ranged from 100 to 26, 578, 118 follows: (1) published in English-language peer-reviewed journals; participants and ages 10 to 86þ years. Females were the focus of (2) contained a statement in the abstract, introduction, or methods 73% of studies, whereas 2% of studies included men only and 25% that the weathering hypothesis was being tested in relation to included both men and women. physical health outcomes; and (3) compared at least one disad- The weathering hypothesis was tested most often for birth vantaged group with one advantaged group. outcomes (58%). Potentially disadvantaged racial or ethnic groups to whom the weathering hypothesis was applied Study selection included Blacks born in the United States, Blacks born in the United Kingdom, Puerto Ricans, Black Hispanics, Mexican Applying the selection criteria to the retrieved articles, one Americans born in the United States, and Taiwan-born Aborigi- reviewer examined the titles and abstracts for full-text review and nals. In comparison, advantaged groups included Whites, White two reviewers independently assessed the eligibility of the full-text Hispanics, foreign-born Mexican Americans, and non-Aboriginals articles, reaching a consensus on the final articles to include in the (Table 1). Black and White racial group comparisons appeared systematic review. most frequently across the studies (80%). Fifty-one percent of the studies included disadvantaged groups other than Blacks, and Data extraction 22% of studies examined weathering by nativity status.

One reviewer used a data collection form (Table A1) to extract Overview of the evidence by type of test of the weathering data from the included articles and a second reviewer veriﬁed the hypothesis extracted data to ensure accuracy. The publication characteristics (author name, publication year), study characteristics (data source), Age patterns participant characteristics (age, gender), weathering hypothesis Comparing the age patterns of health outcomes was used as one characteristics (type of test employed to test the weathering hy- of the ﬁrst approaches to determine the presence of weathering. pothesis), and study characteristics related to potential risks of bias Among the 28 studies that used the age patterns test of weathering (participant selection and retention) were recorded for each article. (Table 2, Table 3), evidence of weathering was more likely to be observed in studies on birth outcomes (37/44 tests) (low birth Risk of bias assessment weight [11/14] [15,20e27], mean birth weight [1/1] [28], moderately low birth weight [5/5] [23,29,30], very low birth weight [6/7] The methodological quality of the included articles was assessed [15,20,23,29], infant mortality [5/5] [20,31e33], neonatal mortality using the Quality in Prognostic Studies (QUIPS) tool [19], which we [2/2] [20,34], postneonatal mortality [1/1] [20], preterm birth [3/5] adapted for this review (Table A2) by dichotomizing studies as low [35e37], small for gestational age [3/3] [21,24,38], and intrauterine (combined original QUIPS criteria for low and moderate risk) or growth retardation [0/1]) and for non-birth outcomes (11/15) high (original QUIPS criteria for high risk) risk and replacing the (body mass index [1/1] [20], diabetes [1/1] [39], hypertension/ term prognostic factor with exposure. blood pressure [4/4] [20,34,39,40], stroke [1/1] [39], functional Two reviewers independently assessed the likelihood of bias limitations [2/3] [41,42], longevity [1/1] [43], self-reported health by examining the extracted study information on potential risk of [1/1] [20], cardiovascular disease [0/1], cholesterol [0/1], and ane- bias (Table A1) and then using the QUIPS (Table A2) “basis for mia [0/1]). judgment” criteria for six bias domains (study participation, study Although weathering age patterns were observed for most attrition, exposure measurement, outcome measurement, con- health outcomes, the results varied by the comparison group founding, and statistical analysis and reporting) as a guide for studied. Among the studies comparing the health of U.S. Blacks to A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18 3

Identification

Articles identified in the citation and database searches: n=817

Duplicates excluded in EndNote: n=200

Articles remaining after removal of duplicates: n=617

Screening

Titles and abstracts screened: n=617

Excluded after title and abstract review: n=276

Eligibility

Full-text articles assessed for eligibility: n=341

Excluded after full-text review: n=301 *no explicit test (n=298): -no test, but conclusion states if weathering could explain the results (n=265) -no test and does not mention if weathering could explain the results (n=33) Included *no comparison group (n=3)

Full-text articles n=40

Articles retrieved from reference lists meeting inclusion criteria: n=1

Full-text articles included in the current analysis: n=41

Fig. 1. Flow diagram of study selection.

Whites by age, all but four studies (on preterm birth [24], choles- considered measures of socioeconomic disadvantage defined by terol level [20], cardiovascular disease [39], and functional limita- individual or neighborhood poverty, education, and neighbor- tions [44]) observed evidence of weathering. Given that this hood segregation. Among these studies, 14/20 tests showed that hypothesis is thought to have relevance for any disadvantaged weathering was more likely to be observed in the socioeco- population, evidence of weathering was also observed among nomically disadvantaged group than the socioeconomically disadvantaged groups, such as Blacks of Caribbean and African advantaged group, and all studies supporting the weathering descent in the United Kingdom [25], Puerto Ricans [23], Mexican hypothesis were for birth outcomes [15,21,24e27,29,30,35e37] Americans [23,31,33,34], and Black Hispanics [42]. However, evi- (Table 2). dence of weathering was not observed for any health outcome among the aggregated Hispanic ethnic group (for low birth weight Biological/physiological mechanisms [22], mean birth weight [28], small for gestational age [38], func- Several studies have investigated different biological/physio- tional limitations [41]) or among Taiwanese Aboriginals (for low logical mechanisms (i.e., allostatic load, telomere length, chronic birth weight [45]). stress, inflammation, and epigenetics) hypothesized to link chronic exposure to social/economic disadvantage to racial or ethnic health Socioeconomic status disparities. Among thirteen studies, eight focused on allostatic load, In addition to studies that defined disadvantage by race or two on telomere length, and one study each on chronic stress, ethnicity, several studies [15,21,24e27,29,30,35e37,46] inflammation, and epigenetics (Table 4). Table 1 4 Characteristics of selected studies explicitly testing the weathering hypothesis

Publication Participant characteristics Study characteristics Weathering characteristics hypothesis characteristics

Author and Age Gender Race/ethnicity/nativity Geographic location Data source Sample size Study design Physical health outcome Evidence of year (in years) weathering

Bird 2010 20e86 Men and Blacks United States NHANES III 13,184 Cross-sectional Allostatic load Yes women Mexican Americans Whites

Borrell 2010 25e65 Men and Blacks United States NHANES III and NDI 13,715 Cohort All-cause mortality Yes women Mexican Americans Whites

Buescher 2006 15e35 Women Blacks North Carolina BRFSS and North Carolina 495,551 Cross-sectional Low birth weight Yes Whites birth and infant death Very low birth weight

records Infant mortality 1 (2019) 33 Epidemiology of Annals / al. et Forde A.T. Neonatal mortality Postneonatal mortality Body mass index High blood pressure High cholesterol Poor self-reported health

Chinn 2016 18e85 Women U.S.-born Black Hispanics United States NHIS 42,908 Cross-sectional Functional limitations Yes Foreign-born Black Hispanics U.S.-born White Hispanics Foreign-born White Hispanics U.S.-born other-race Hispanics Foreign-born other-race Hispanics

Chyu 2011 18e70 Women Blacks United States NHANES IV 5765 Cross-sectional Allostatic load Yes U.S.-born Mexican Americans

Foreign-born Mexican Americans e 18 Whites

Cohen 2016 12e40 Women Blacks United States NCHS 2,960,578 Cross-sectional Infant mortality Yes Mexican Americans Whites

Collins 2006 <20e40 Women Extremely impoverished Blacks Illinois Illinois Vital Records, U.S. 46,725 Cross-sectional Very low birth weight Yes Nonimpoverished Blacks Census and Chicago Moderately low birth Department of Public weight Health data

Collins 2009 14e35 Women Blacks Illinois Illinois transgenerational 267,303 Cross-sectional Low birth weight Yes Whites dataset and U.S. Census (Stratiﬁed by early-life/adulthood residence income: low/low, low/ high, high/low, high/high)

Collins 2012 <20e35 Women U.S.-born Mexican Americans Illinois Illinois transgenerational 267,303 Cross-sectional Low birth weight No Foreign-born Mexican Americans dataset and U.S. Census Preterm birth Intrauterine growth retardation (IUGR) Collins 2015 14e35 Women Former low birth weight Blacks Illinois Illinois transgenerational 70,580 Cross-sectional Low birth weight Yes Former low birth weight Whites dataset and U.S. Census Small for gestational age Nonelow birth weight Blacks Nonelow birth weight Whites

Das 2013 57e85 Men Blacks United States NSHAP 1455 Cross-sectional Metabolic outcomes Yes Whites (blood sugar/diabetes, blood pressure, heart rate)

Dennis 2013 15e35 Women Blacks United States ECLS-B 6150 Cross-sectional Low birth weight Yes U.S.-born Hispanics Foreign-born Hispanics Whites

Geronimus 1996 15e34 Women Blacks Michigan Michigan Vital Records and 54,888 Cross-sectional Low birth weight Yes Whites U.S. Census Very low birth weight

Geronimus 2006 18e64 Men and Blacks United States NHANES IV 6586 Cross-sectional Allostatic load Yes women Whites ..Free l naso pdmooy3 21)1 (2019) 33 Epidemiology of Annals / al. et Forde A.T. Geronimus 2007 15e65 Men and Blacks United States NHANES IV 5501 Cross-sectional Hypertension Yes women Whites

Geronimus 2010 49e55 Women Blacks Boston, Chicago, SWAN 215 Cohort Telomere length Yes Whites Detroit, Oakland, Los Angeles, Newark, Pittsburgh

Geronimus 2015 25e75þ Men and Blacks Detroit HEP Community Survey 202 Cross-sectional Telomere length Yes women Mexican Americans Whites

Goisis 2014 14e30þ Women U.K. Blacks United Kingdom ONSLS 45,856 Cross-sectional Low birth weight Yes U.K. Whites (Stratiﬁed by area-level advantage/ disadvantage, high/low education) e Hibbs 2016 14e35 Women Black smokers and non-smokers Illinois Illinois transgenerational 267,303 Cross-sectional Preterm birth Yes 18 White smokers and non-smokers dataset and U.S. Census (Stratiﬁed by quartiles of neighborhood income) 182,938 Cross-sectional Preterm birth Yes

e Holzman 2009 20 39 Women Black smokers and non-smokers Philadelphia, Baltimore, The MultilevelExplaining Modeling Preterm of White smokers and non-smokers 16 Michigan cities, DisparitiesDelivery Project and U.S. ﬁ (Strati ed by low/medium/high 3 Maryland counties, census neighborhood deprivation) 2 North Carolina Howard 2016 25þ Men and Blacks countiesUnited States NHANES III and NDI 11,733 Cross-sectional Mortality Yes women Mexican Americans Whites

Kaestner 2009 30e60 Men and Blacks United States NHANES III 7010 Cross-sectional Allostatic load Yes women U.S.-born Mexican Americans Foreign-born Mexican Americans Whites

(continued on next page) 5 Table 1 (continued ) 6

Publication Participant characteristics Study characteristics Weathering characteristics hypothesis characteristics

Author and Age Gender Race/ethnicity/nativity Geographic location Data source Sample size Study design Physical health outcome Evidence of year (in years) weathering

Khoshnood 2005 20e35þ Women Blacks United States NCHS 8,433,935 Cross-sectional Low birth weight Yes Mexican Americans Moderately low birth Puerto Ricans weight Whites Very low birth weight

Kramer 2014 10e35þ Women Blacks Georgia Georgia birth records and 1,000,437 Cohort Preterm-low birth weight Yes Hispanics U.S. Census Whites (Stratiﬁed by low and high neighborhood deprivation)

Lin 2017 55e65 Men and Blacks United States HRS 7715 Cohort Functional limitations Yes

women Hispanics 1 (2019) 33 Epidemiology of Annals / al. et Forde A.T. Whites

Love 2010 <20e35 Women Blacks Illinois Illinois transgenerational 70,615 Cross-sectional Low birth weight Yes Whites dataset and U.S. Census Small for gestational age (Stratiﬁed by early/adult Preterm birth neighborhood income: lower- lower, upper-lower, lower-upper, upper-upper lifetime)

Osypuk 2008 15e45 Women Blacks United States NCHS and U.S. Census 1,944,703 Cross-sectional Preterm birth Yes Whites (Stratiﬁed by segregation status: nonhypersegregated and hypersegregated metropolitan area)

Peek 2010 25e65þ Men and Blacks Texas City Texas City Stress and Health 1410 Cross-sectional Allostatic load Yes e 18 women U.S.-born Mexican Americans Study Foreign-born Mexican Americans Whites

Powers 2013 <20e40þ Women Blacks United States NCHS 26,578,118 Cross-sectional Infant mortality Yes U.S.-born Mexican Americans Foreign-born Mexican Americans Whites

Powers 2016 25e40þ Women U.S.-born Mexican Americans United States NCHS 14,542,120 Cross-sectional Infant mortality Yes Foreign-born Mexican Americans Whites New York City Bureau of vital statistics, 158,174 Cross-sectional Moderately low birth Yes New York City Department weight Rauh 2001 20e39 Women Blacks of Health and U.S. Census Very low birth weight Whites Sheeder 2006 18e34 Women Blacks Colorado Colorado birth certiﬁcate 91,061 Cross-sectional Small for gestational age Yes Hispanics data Whites

Simons 2016 Mean Women Blacks Georgia and Iowa FACHS 100 Cohort Epigenetic measure of Yes ¼48.5 biological aging/ methylation Spence 2008 65e83 Women Blacks United States NLS-MW 1608 Cohort Functional limitations No Whites

Spence 2009 45e59 Women Blacks United States NLS-MW 3769 Cohort Longevity (post Yes Whites reproductive mortality)

Strutz 2014 11e32 Women Blacks United States Add Health 5413 Cohort Birth weight Yes Mexican Americans Non-Mexican Latinas Whites

Swamy 2012 15e44 Women Blacks North Carolina North Carolina birth record 510,288 Cross-sectional Mean birth weight Yes Hispanics database Whites

Thorpe 2016 18e75þ Men and Blacks United States NHIS 619,130 Cross-sectional Hypertension Yes women Whites Stroke Diabetes Cardiovascular Disease ..Free l naso pdmooy3 21)1 (2019) 33 Epidemiology of Annals / al. et Forde A.T. Wallace 2013 20e35 Women Blacks New Orleans Tulane-Lakeside Hospital 123 Cohort Birth weight Yes Whites Department of Obstetrics Birth weight ratio and Gynecology Gestational age Birth length Head circumference

Wang 2012 15e35 Women Aboriginals Taiwan Taiwan birth registration 8432 Cross-sectional Low birth weight or No Non-Aboriginals database preterm birth

Wildsmith 2002 15e34 Women U.S.-born Mexican Americans United States NCHS 387,909 Cross-sectional Low birth weight Yes Foreign-born Mexican Americans Anemia Hypertension Neonatal mortality

Add Health ¼ National Longitudinal Study of Adolescent Health; BRFSS ¼ Behavioral Risk Factor Surveillance System; ECLSB ¼ Early Childhood Longitudinal Study-Birth Cohort; FACHS ¼ Family and Community Health Study; HEP ¼ The Healthy Environment Partnership; HRS ¼ Health and Retirement Study; NCHS ¼ National Center for Health Statistics; NDI ¼ National Death Index; NHANES ¼ National Health and Nutrition Examination Survey; NHIS ¼ National Health Interview Survey; NLS-MW ¼ National Longitudinal Survey of Mature Women; NSHAP ¼ National Social Life, Health and Aging Project; ONSLS ¼ Ofﬁce for National Statistics longitudinal study; SWAN ¼ ¼ ¼ e Study of Women's Health Across the Nation; U.K. United Kingdom; U.S. United States. 18 7 Table 2 8 Summary of the evidence among studies supporting the weathering hypothesis * Author and Health outcome Population for which Comparison population Weathering patterns by socioeconomic status publication year weathering was hypothesized to have the greatest impact

Birth outcomes

Buescher 2006 Low birth weight Blacks Whites N/A

Collins 2009 Low birth weight N/A N/A Weathering more pronounced in Blacks with lifelong residence in lower SES neighborhoods

Collins 2015 Low birth weight Former non-low Whites, former low birth weight Blacks Weathering more pronounced in non-low birth weight Blacks in birth weight Blacks lower SES neighborhoods

Dennis 2013 Low birth weight Blacks Whites, Hispanics (U.S.-born or foreign-born) N/A

Geronimus 1996 Low birth weight Blacks Whites Weathering more pronounced in lower SES Blacks ..Free l naso pdmooy3 21)1 (2019) 33 Epidemiology of Annals / al. et Forde A.T. Goisis 2014 Low birth weight U.K. Blacks U.K. Whites Weathering more pronounced in less-educated Blacks and Blacks in lower SES neighborhoods

Khoshnood 2005 Low birth weight Blacks Whites N/A

Khoshnood 2005 Low birth weight Mexican Americans Whites N/A

Khoshnood 2005 Low birth weight Puerto Ricans Whites N/A

Kramer 2014 Preterm-low birth weight N/A N/A Weathering more pronounced in Blacks with higher neighborhood deprivation

Love 2010 Low birth weight Blacks Whites Weathering more pronounced in Blacks with lifelong residence in lower SES neighborhoods

Swamy 2012 Mean birth weight Blacks Whites, Hispanics N/A e 18 Collins 2006 Moderately low birth weight N/A N/A Weathering more pronounced in Blacks living in lower SES neighborhoods

Khoshnood 2005 Moderately low birth weight Blacks Whites N/A

Khoshnood 2005 Moderately low birth weight Mexican Americans Whites N/A

Khoshnood 2005 Moderately low birth weight Puerto Ricans Whites N/A

Rauh 2001 Moderately low birth weight Blacks Whites Weathering more pronounced in lower SES Blacks

Buescher 2006 Very low birth weight Blacks Whites N/A

Geronimus 1996 Very low birth weight Blacks Whites Weathering more pronounced in lower SES Blacks

Khoshnood 2005 Very low birth weight Blacks Whites N/A

Khoshnood 2005 Very low birth weight Mexican Americans Whites N/A

Khoshnood 2005 Very low birth weight Puerto Ricans Whites N/A

Rauh 2001 Very low birth weight Blacks Whites No difference by SES

Buescher 2006 Infant mortality Blacks Whites N/A Cohen 2016 Infant mortality Blacks Whites, Mexican Americans N/A

Powers 2013 Infant mortality Blacks Whites N/A

Powers 2013 Infant mortality Mexican Americans Whites N/A

Powers 2016 Infant mortality Mexican Americans Whites N/A

Buescher 2006 Neonatal mortality Blacks Whites N/A

Wildsmith 2002 Neonatal mortality U.S.-born Mexican Americans Foreign-born Mexican Americans N/A

Buescher 2006 Postneonatal mortality Blacks Whites N/A

Hibbs 2016 Preterm birth N/A N/A Weathering more pronounced among cigarette smoking Blacks with early-life or lifelong residence in lower SES neighborhoods

Holzman 2009 Preterm birth Black smokers, Black non- White non-smokers Weathering more pronounced in Black and White smokers with smokers, White smokers higher neighborhood deprivation ..Free l naso pdmooy3 21)1 (2019) 33 Epidemiology of Annals / al. et Forde A.T.

Osypuk 2008 Preterm birth Blacks Whites Weathering more pronounced for Blacks in hypersegregated neighborhoods

Collins 2015 Small for gestational age Former nonelow birth weight Whites, former low birth weight Blacks Weathering more pronounced in former nonelow birth weight Blacks Blacks in lower SES neighborhoods

Love 2010 Small for gestational age Blacks Whites Weathering more pronounced in Blacks with lifelong residence in lower SES neighborhoods

Sheeder 2006 Small for gestational age Blacks Whites, Hispanics N/A

Nonebirth outcomes

Buescher 2006 Body mass index Blacks Whites N/A

Thorpe 2016 Diabetes Blacks Whites N/A e 18

Buescher 2006 High blood pressure Blacks Whites N/A

Geronimus 2007 Hypertension Blacks Whites N/A

Thorpe 2016 Hypertension Blacks Whites N/A

Wildsmith 2002 Hypertension U.S.-born Mexican Americans Foreign-born Mexican Americans N/A

Thorpe 2016 Stroke Blacks Whites N/A

Chinn 2016 Functional limitations U.S.-born Black Hispanics Whites, Other-race Hispanics (U.S. and foreign- N/A born), Foreign-born Black Hispanics

Lin 2017 Functional limitations Blacks Whites, Hispanics N/A

Spence 2009 Longevity Blacks Whites N/A

* BuescherRepresents 2006 studies for Poor which self-reported a group comparison health for testing Blacks the weathering hypothesis was deﬁned Whites by socioeconomic status (SES) alone or in addition to N/A a comparison deﬁned by membership in a historically disadvantaged N/Aminority¼ not group applicable; versus U.K. an advantaged¼ United Kindgom; group. U.S. ¼ United States. 9 10 A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18

Table 3 Summary of the evidence among studies not supporting the weathering hypothesis

Author and Health outcome Population for which Comparison population Weathering patterns * publication year weathering was hypothesized by socioeconomic status to have the greatest impact

Birth outcomes

Collins 2012 Low birth weight U.S.-born Mexican Americans Foreign-born Mexican No difference by neighborhood SES Americans

Wildsmith 2002 Low birth weight U.S.-born Mexican Americans Foreign-born Mexican N/A Americans

Wang 2013 Low birth weight or preterm birth Aboriginal Non-Aboriginal N/A

Collins 2006 Very low birth weight N/A N/A No difference by neighborhood SES

Collins 2012 Intrauterine growth retardation U.S.-born Mexican Americans Foreign-born Mexican No difference by neighborhood SES Americans Other birth outcomes

Collins 2012 Preterm birth U.S.-born Mexican Americans Foreign-born Mexican No difference by neighborhood SES Americans

Love 2010 Preterm birth Blacks Whites No difference by neighborhood SES

Nonebirth outcomes

Thorpe 2016 Cardiovascular disease Blacks Whites N/A

Buescher 2006 High cholesterol Blacks Whites N/A

Spence 2008 Functional limitations Blacks Whites N/A

Wildsmith 2002 Anemia U.S.-born Mexican Americans Foreign-born Mexican N/A Americans

N/A ¼ not applicable; U.S. ¼ United States. * Represents studies for which a group comparison for testing the weathering hypothesis was deﬁned by socioeconomic status (SES) alone, or in addition to a comparison deﬁned by membership in a historically disadvantaged minority group versus an advantaged group.

Five studies [47e51] included allostatic load as the primary were potential threats to validity most likely arising from selection health outcome and reported results that were supportive of bias. weathering when considering either race or ethnicity [47e51] or socioeconomic status (SES) [47,51] as measures of disadvantage Discussion (Table 4). Three additional studies [52e54] examining allostatic load as the primary exposure also showed evidence of weathering, The weathering hypothesis has been tested for several health where higher allostatic load was associated with increased mor- outcomes among a diverse group of participants. Most studies tality rates [52,53] and lower gestational age [54] (Table 4). focused on birth outcomes for Blacks versus Whites, consistent Studies on telomere length, chronic stress, inflammation, and with the original framing of the weathering hypothesis. Generally, epigenetics also found support for weathering. Regarding telomere however, findings supported the weathering hypothesis for both length as a marker of biological age, one study [55] reported that birth and non-birth outcomes. The most common approach used to Black women had shorter telomere length than White women and test the weathering hypothesis was to compare disadvantaged to the difference was partially explained by perceived stress and advantaged groups across different ages to determine if age-related poverty. Another study [56] showed decreasing telomere length patterns of adverse health outcomes were accelerated among with age and observed an interaction between poverty and race or disadvantaged versus advantaged groups. Fewer studies focused on ethnicity. Chronic stress was associated with lower birth weight in biological mechanisms explaining these patterns or how SES may Blacks, Mexican-origin Latinas, and noneMexican origin Latinas exacerbate the relationship between age and poor health. compared with Whites [57].Inflammation arising from cumulative This review identified a diverse literature on the weathering exposure to stress placed Black men at a greater risk for developing hypothesis that was determined to be of good methodological diabetes and cardiovascular disease than White men [58]. Having quality, although there was some concern about the potential for lower income versus higher income resulted in accelerated aging selection bias and the temporal ambiguity. The cross-sectional (as measured by methylation changes) among Black women [59] nature of the data prevented the longitudinal examination of (Table 4). health outcomes over the life course and introduced potential se- Overall, most studies provided evidence in support of weath- lection bias. Specifically, prevalence-incidence bias or selective ering that was more pronounced in racial or ethnic minority survival bias was a possibility because a disproportionately high groups, lower SES groups, and segregated neighborhoods, but re- percentage of cases with long duration (prevalent cases) and better sults for studies including biological mechanisms were not neces- average survival could have been included in cross-sectional sarily restricted to racial or ethnic minority groups (Table 4). Most studies, whereas those with shorter duration (incident cases) studies were of good methodological quality (Table 5), but there who represented the complete range of severity of the health A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18 11

Table 4 Summary of the evidence for studies exploring biological/physiological mechanisms underlying the weathering hypothesis

Author and publication year Health outcome Evidence of weathering in the overall population

Birth outcomes

Strutz 2014 Birth weight Chronic stress was associated with low birth weight

Wallace 2013 Birth length e

Wallace 2013 Birth weight e

Wallace 2013 Birth weight ratio e

Wallace 2013 Head circumference e

Wallace 2013 Gestational age Increasing allostatic load was associated with decreasing gestational age No racial variation in the association

Nonebirth outcomes

Bird 2010 Allostatic load Blacks had the highest allostatic load Living in a lower SES neighborhood was associated with higher allostatic load, regardless of race and ethnicity

Chyu 2011 Allostatic load Allostatic load was higher for Blacks at earlier ages (more weathered) than Whites Allostatic load was higher for U.S.-born Mexican Americans than foreign-born Mexican Americans

Geronimus 2006 Allostatic load Allostatic load was higher for Blacks at earlier ages (more weathered) than Whites Weathering was more pronounced for those with lower SES than higher SES

Kaestner 2009 Allostatic load Allostatic load was higher for Blacks than Whites Allostatic load was higher for U.S.-born Mexican Americans than foreign-born Mexican Americans

Peek 2010 Allostatic load Blacks had the highest allostatic load U.S.-born Mexican Americans had higher allostatic load than foreign-born Mexican Americans

Das 2013 Metabolic outcomes Inﬂammation was associated with an increased risk for diabetes and cardiovascular disease The association was stronger for Blacks than Whites

Simons 2016 Epigenetic measure/methylation Lower income was associated with accelerated aging

Geronimus 2010 Telomere length Telomere length was shorter for Blacks (more weathered) than Whites

Geronimus 2015 Telomere length Telomere length decreased with age No racial/ethnic variation in telomere length for Blacks, Mexican Americans, and Whites Weathering was more pronounced in lower SES Whites than higher SES Whites

Borrell 2010 All-cause mortality Higher allostatic load ( 3) was associated with increased mortality risk compared with lower allostatic load ( 1) No racial and ethnic variation in the association

Howard 2016 Mortality Increasing allostatic load was associated with increased mortality risk The association was stronger for Blacks than Whites e ¼ no evidence of weathering; U.S. ¼ United States; SES ¼ socioeconomic status. outcome would have a lower probability of being included. The reliable measurements (e.g., self-reported outcomes that would studies that examined short-acting health conditions were less only be reported if notified by a physician). Confounding was also likely to be impacted by this bias. Among the cohort studies, there less likely because most studies included valid and reliable appeared to be adequate response rates and attempts to collect measurement and analysis of confounders that were clearly information on participants who were not included in the studies, defined. but studies with significant differences in participation and attri- Some important limitations of our review should be noted. tion were judged as being more likely to have appreciable selection Relevant studies that may have inexplicitly tested the weathering bias that impacted the results. In contrast, although temporality in hypothesis, but were omitted because weathering or Geronimus' cross-sectional studies is often difficult to establish, temporal seminal articles were not specified, were likely relevant to a inference in the context of weathering is less problematic as comprehensive review of the weathering hypothesis. The presence weathering is presumed to begin at birth (or even in utero) and of publication bias was likely, but there was not sufficient infor- therefore precede the health outcome. mation to assess publication bias. Although measurement error was less likely, the few studies The weathering hypothesis has contributed significantly to the rated as high risk for measurement error failed to use valid and literature on racial disparities in birth outcomes. The number of Table 5 12 Potential risk of bias assessment for the included studies

Author and Evidence of Health outcome Geographic Dataset Sample Study Study Study Exposure Outcome Study Statistical Overall publication weathering location size design participation attrition measurement measurement confounding analysis potential year and for bias reporting

Geronimus 2010 Yes Telomere length Boston, Chicago, SWAN 215 Cohort High High Low Low Low Low High Detroit, Oakland, Los Angeles, Newark, Pittsburgh

Simons 2016 Yes Epigenetic measure Georgia and Iowa FACHS 100 Cohort High High Low Low Low Low High of biological aging/ methylation

Wallace 2013 Yes Gestational age Louisiana Tulane-Lakeside 123 Cohort High High Low Low Low Low High Hospital Department of Obstetrics and Gynecology 1 (2019) 33 Epidemiology of Annals / al. et Forde A.T.

Wallace 2013 No Birth weight Louisiana Tulane-Lakeside 123 Cohort High High Low Low Low Low High Hospital Department of Obstetrics and Gynecology

Wallace 2013 No Birth weight ratio Louisiana Tulane-Lakeside 123 Cohort High High Low Low Low Low High Hospital Department of Obstetrics and Gynecology

Wallace 2013 No Birth length Louisiana Tulane-Lakeside 123 Cohort High High Low Low Low Low High Hospital Department of Obstetrics and Gynecology e

Wallace 2013 No Head Louisiana Tulane-Lakeside 123 Cohort High High Low Low Low Low High 18 circumference Hospital Department of Obstetrics and Gynecology

Spence 2008 No Functional United States NLS-MW 1608 Cohort Low High Low Low Low Low High limitations

Borrell 2010 Yes All-cause mortality United States NHANES III and NDI 13,715 Cohort Low High Low Low Low Low High

Kramer 2014 Yes Preterm low birth Georgia Georgia birth records 1,000,437 Cohort Low High Low Low Low Low High weight and U.S. Census

Lin 2017 Yes Functional United States HRS 7715 Cohort Low High Low Low Low Low High limitations

Spence 2009 Yes Longevity (post United States NLS-MW 3769 Cohort Low High Low Low Low Low High reproductive mortality)

Strutz 2014 Yes Birth weight United States Add Health 5413 Cohort Low High Low Low Low Low High

Geronimus 2015 Yes Telomere length Detroit HEP Community 202 Cross- High Low Low Low Low Low High Survey sectional Buescher 2006 Yes High blood North Carolina BRFSS and North 495,551 Cross- Low Low Low High High Low High pressure Carolina birth and sectional infant death records

Buescher 2006 Yes Body mass index North Carolina Behavioral BRFSS and 495,551 Cross- Low Low Low High High Low High North Carolina birth sectional and infant death records

Buescher 2006 No Cholesterol North Carolina BRFSS and North 495,551 Cross- Low Low Low High High Low High Carolina birth and sectional infant death records

Wildsmith 2002 Yes Hypertension United States NCHS 387, 909 Cross- Low Low Low High High Low High sectional

Wildsmith 2002 No Anemia United States NCHS 387,909 Cross- Low Low Low High High Low High sectional

Thorpe 2016 Yes Diabetes United States NHIS 619,130 Cross- Low Low Low High Low Low High ..Free l naso pdmooy3 21)1 (2019) 33 Epidemiology of Annals / al. et Forde A.T. sectional

Thorpe 2016 Yes Hypertension United States NHIS 619,130 Cross- Low Low Low High Low Low High sectional

Thorpe 2016 Yes Stroke United States NHIS 619,130 Cross- Low Low Low High Low Low High sectional

Thorpe 2016 No Cardiovascular United States NHIS 619,130 Cross- Low Low Low High Low Low High disease sectional

Buescher 2006 Yes Infant mortality North Carolina BRFSS and North 495,551 Cross- Low Low Low Low High Low High Carolina birth and sectional infant death records

Buescher 2006 Yes Low birth weight North Carolina BRFSS and North 495,551 Cross- Low Low Low Low High Low High

Carolina birth and sectional e infant death records 18

Buescher 2006 Yes Neonatal mortality North Carolina BRFSS and North 495,551 Cross- Low Low Low Low High Low High Carolina birth and sectional infant death records

Buescher 2006 Yes Postneonatal North Carolina Behavioral Risk 495,551 Cross- Low Low Low Low High Low High mortality Factor Surveillance sectional System (BRFSS) and North Carolina birth & infant death records

Buescher 2006 Yes Self-reported North Carolina BRFSS and North 495,551 Cross- Low Low Low Low High Low High health Carolina birth and sectional infant death records

Buescher 2006 Yes Very low birth North Carolina BRFSS and North 495,551 Cross- Low Low Low Low High Low High weight Carolina birth and sectional infant death records 13

Wildsmith 2002 Yes Neonatal mortality United States NCHS 387, 909 Cross- Low Low Low Low High Low High sectional (continued on next page) Table 5 (continued ) 14

Wildsmith 2002 No Low birth weight United States NCHS 387,909 Cross- Low Low Low Low High Low High sectional

Bird 2010 Yes Allostatic load United States NHANES III 13,184 Cross- Low Low Low Low Low Low Low sectional

Chinn 2016 Yes Functional United States NHIS 42,908 Cross- Low Low Low Low Low Low Low limitations sectional

Collins 2006 No Very low birth Illinois Illinois Vital Records, 46,725 Cross- Low Low Low Low Low Low Low weight U.S. Census and sectional Chicago Department of Public Health data

Chyu 2011 Yes Allostatic load United States NHANES IV 5765 Cross- Low Low Low Low Low Low Low 1 (2019) 33 Epidemiology of Annals / al. et Forde A.T. sectional

Cohen 2016 Yes Infant mortality United States NCHS 2,960,578 Cross- Low Low Low Low Low Low Low sectional

Collins 2006 Yes Moderately low Illinois Illinois Vital Records, 46,725 Cross- Low Low Low Low Low Low Low birth weight U.S. Census & Chicago sectional Department of Public Health data

Collins 2009 Yes Low birth weight Illinois Illinois 267,303 Cross- Low Low Low Low Low Low Low transgenerational sectional dataset and U.S. Census

Collins 2012 No Low birth weight Illinois Illinois 267,303 Cross- Low Low Low Low Low Low Low e

transgenerational sectional 18 dataset and U.S. Census

Collins 2012 No Preterm birth Illinois Illinois 267,303 Cross- Low Low Low Low Low Low Low transgenerational sectional dataset and U.S. Census

Collins 2012 No Intrauterine Illinois Illinois 267,303 Cross- Low Low Low Low Low Low Low growth retardation transgenerational sectional dataset and U.S. Census

Collins 2015 Yes Low birth weight Illinois Illinois 70,580 Cross- Low Low Low Low Low Low Low transgenerational sectional dataset and U.S. Census

Collins 2015 Yes Small for Illinois Illinois 70,580 Cross- Low Low Low Low Low Low Low gestational age transgenerational sectional dataset and U.S. Census Das 2013 Yes Metabolic United States NSHAP 1455 Cross- Low Low Low Low Low Low Low outcomes (blood sectional sugar/diabetes, blood pressure, heart rate)

Dennis 2013 Yes Low birth weight United States ECLS-B 6150 Cross- Low Low Low Low Low Low Low sectional

Geronimus 1996 Yes Low birth weight Michigan Michigan Vital 54,888 Cross- Low Low Low Low Low Low Low Records and U.S. sectional Census

Geronimus 1996 Yes Very low birth Michigan Michigan Vital 54,888 Cross- Low Low Low Low Low Low Low weight Records and U.S. sectional Census

Geronimus 2006 Yes Allostatic load United States NHANES IV 6586 Cross- Low Low Low Low Low Low Low sectional ..Free l naso pdmooy3 21)1 (2019) 33 Epidemiology of Annals / al. et Forde A.T. Geronimus 2007 Yes Hypertension United States NHANES IV 5501 Cross- Low Low Low Low Low Low Low sectional

Goisis 2014 Yes Low birth weight United Kingdom ONSLS 45,856 Cross- Low Low Low Low Low Low Low sectional

Hibbs 2016 Yes Preterm birth Illinois Illinois 267,303 Cross- Low Low Low Low Low Low Low transgenerational sectional dataset and U.S. Census

Holzman 2009 Yes Preterm birth Philadelphia, The Multilevel 182,938 Cross- Low Low Low Low Low Low Low Baltimore, 16 Modeling of sectional Michigan cities, 3 Disparities Explaining Maryland counties, Preterm Delivery 2 North Carolina Project and U.S.

counties census e 18

Howard 2016 Yes Mortality United States NHANES III and NDI 11,733 Cross- Low Low Low Low Low Low Low sectional

Kaestner 2009 Yes Allostatic load United States NHANES III 7010 Cross- Low Low Low Low Low Low Low sectional

Khoshnood 2005 Yes Low birth weight United States NCHS 8,433,935 Cross- Low Low Low Low Low Low Low sectional

Khoshnood 2005 Yes Moderately low United States NCHS 8,433,935 Cross- Low Low Low Low Low Low Low birth weight sectional

Khoshnood 2005 Yes Very low birth United States NCHS 8,433,935 Cross- Low Low Low Low Low Low Low weight sectional

Love 2010 Yes Low birth weight Illinois Illinois 70,615 Cross- Low Low Low Low Low Low Low transgenerational sectional dataset and U.S. Census 15 (continued on next page) Table 5 (continued ) 16

Love 2010 Yes Small for Illinois Illinois 70,615 Cross- Low Low Low Low Low Low Low gestational age transgenerational sectional dataset and U.S. Census

Love 2010 No Preterm birth Illinois Illinois 70,615 Cross- Low Low Low Low Low Low Low transgenerational sectional dataset and U.S. Census

Osypuk 2008 Yes Preterm birth United States NCHS and U.S. Census 1,944,703 Cross- Low Low Low Low Low Low Low sectional

Peek 2010 Yes Allostatic load Texas City Texas City Stress and 1410 Cross- Low Low Low Low Low Low Low Health Study sectional 1 (2019) 33 Epidemiology of Annals / al. et Forde A.T.

Powers 2013 Yes Infant mortality United States NCHS 26,578,118 Cross- Low Low Low Low Low Low Low sectional

Powers 2016 Yes Infant mortality United States NCHS 14,542,120 Cross- Low Low Low Low Low Low Low sectional

Rauh 2001 Yes Moderately low New York City Bureau of vital 158,174 Cross- Low Low Low Low Low Low Low birth weight statistics, New York sectional City Department of Health and U.S. Census

Rauh 2001 Yes Very low birth New York City Bureau of vital 158,174 Cross- Low Low Low Low Low Low Low weight statistics, New York sectional City Department of e

Health and U.S. 18 Census

Sheeder 2006 Yes Small for Colorado Colorado birth 91,061 Cross- Low Low Low Low Low Low Low gestational age certiﬁcate data sectional

Swamy 2012 Yes Mean birth weight North Carolina North Carolina birth 510,288 Cross- Low Low Low Low Low Low Low record database sectional

Wang 2012 No Low birth weight or Taiwan Taiwan birth 8432 Cross- Low Low Low Low Low Low Low preterm birth registration database sectional Family and Community Health Study; HEP ¼ The Healthy Environment Partnership; HRS ¼ Health and Retirement Study; NCHS ¼ National Center for Health Statistics; NDI ¼ National Death Index; NHANES ¼ National Health and Nutrition Examination Survey; NHIS ¼ National Health Interview Survey; NLS-MW ¼ National Add HealthLongitudinal¼ National Survey Longitudinal of Mature StudyWomen; of Adolescent NSHAP ¼ National Health; BRFSS Social¼ Life,Behavioral Health and Risk Aging Factor Project; Surveillance ONSLS System;¼ Office FACHS for National¼ Statistics Longitudinal Study; SWAN ¼ Study of Women's Health Across the Nation; U.S.¼ United States. A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18 17 studies that tested the applicability of this hypothesis to other [16] Geronimus AT, Andersen HF, Bound J. Differences in hypertension prevalence outcomes and racial or ethnic groups has grown since the publi- among United-States black-and-white women of childbearing age. Public Health Rep 1991;106(4):393e9. cation of Geronimus' earliest studies on weathering and birth [17] Geronimus AT. On teenage childbearing and neonatal-mortality in the United- outcomes. However, few studies examined weathering within the States. Popul Dev Rev 1987;13(2):245e79. Hispanic population by race or within the Black population by [18] Moher D, Liberati A, Tetzlaff J, Altman DG, PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Int J nativity status, making this an important target for future research Surg 2010;8(5):336e41. on weathering. Although discrimination was not examined in any [19] Hayden JA, van der Windt DA, Cartwright JL, Cot^ e P, Bombardier C. Assessing of the studies, it may be experienced differently by Blacks than bias in studies of prognostic factors. Ann Intern Med 2013;158(4):280e6. [20] Buescher PA, Mittal M. Racial disparities in birth outcomes increase with other racial and ethnic minority groups and therefore could explain maternal age: recent data from North Carolina. N C Med J 2006;67(1): racial or ethnic differences in weathering patterns. The rigor of 16e20. future research can be enhanced by (1) studying clinical disease [21] Collins JW, Rankin KM, Hibbs S. The maternal age related patterns of infant low birth weight rates among non-latino whites and african-americans: The outcomes using gold standard outcome assessments (rather than effect of maternal birth weight and neighborhood income. Matern child participant self-report), (2) including additional health outcomes Health J 2015;19(4):739e44. (e.g., cancer, dementia, pulmonary disease), (3) conducting ana- [22] Dennis JA, Mollborn S. Young maternal age and low birth weight risk: an lyses on the health of ethnic groups by race and on racial groups by exploration of racial/ethnic disparities in the birth outcomes of mothers in the United States. Soc Sci J 2013;50(4):625e34. nativity status, (4) characterizing the precise nature of social and [23] Khoshnood B, Wall S, Lee KS. Risk of low birth weight associated with economic disadvantage of most relevance (e.g., racial discrimina- advanced maternal age among four ethnic groups in the United States. Matern e tion, neighborhood effects, poverty), and (5) using longitudinal Child Health J 2005;9(1):3 9. [24] Love C, David RJ, Rankin KM, Collins JW. Exploring weathering: effects of studies beginning early in the life course to better characterize how lifelong economic environment and maternal age on low birth weight, small disadvantage and health coevolve over the life course. for gestational age, and preterm birth in African-American and white women. Am J Epidemiol 2010;172(2):127e34. [25] Goisis A, Sigle-Rushton W. Childbearing postponement and child well-being: Acknowledgments a complex and varied relationship? Demography 2014;51(5):1821e41. [26] Kramer MR, Dunlop AL, Hogue CJ. Measuring women's cumulative neighborhood deprivation exposure using longitudinally linked vital records: a The authors would like to show their gratitude to Sharon method for life course MCH research. Matern Child Health J 2014;18(2): Schwartz, Ph.D., who provided many helpful insights during this 478e87. review. This work was supported by the National Heart, Lung, and [27] Collins JW, Wambach J, David RJ, Rankin KM. Women's lifelong exposure to Blood Institute of the National Institutes of Health (grant number neighborhood poverty and low birth weight: a population-based study. Matern Child Health J 2009;13(3):326e33. F31HL117613). [28] Swamy GK, Edwards S, Gelfand A, James SA, Miranda ML. Maternal age, birth order, and race: differential effects on birthweight. J Epidemiol Community Health 2012;66(2):136e42. Disclaimer [29] Rauh VA, Andrews HF, Garfinkel RS. The contribution of maternal age to racial disparities in birthweight: a multilevel perspective. Am J Public Health The content is solely the responsibility of the authors and does 2001;91(11):1815e24. fi [30] Collins JW, Simon DM, Jackson TA, Drolet A. Advancing maternal age and not necessarily represent the of cial views of the National In- infant birth weight among urban African Americans: The effect of neighbor- stitutes of Health. hood poverty. Ethn Dis 2006;16(1):180e6. [31] Powers DA. Paradox revisited: a further investigation of racial/ethnic differences in infant mortality by maternal age. Demography 2013;50(2):495e520. References [32] Cohen PN. Maternal age and infant mortality for white, black, and Mexican mothers in the United States. Sociological Sci 2016;3:32e8. [1] Mensah GA, Mokdad AH, Ford ES, Greenlund KJ, Croft JB. State of disparities in [33] Powers DA. Erosion of advantage: decomposing differences in infant mortality cardiovascular health in the United States. Circulation 2005;111(10): rates among older non-hispanic white and Mexican-origin mothers. Popul Res 1233e41. Policy Rev 2016;35(1):23e48. [2] Mokdad AH, Ford ES, Bowman BA, Nelson DE, Engelgau MM, Vinicor F, et al. [34] Wildsmith EM. Testing the weathering hypothesis among Mexican-origin The continuing increase of diabetes in the US. Diabetes Care 2001;24(2):412. women. Ethn Dis 2002;12(4):470e9. [3] Flegal KM, Carroll MD, Ogden CL, Curtin LR. Prevalence and trends in obesity [35] Holzman C, Eyster J, Kleyn M, Messer LC, Kaufman JS, Laraia BA, et al. Maternal among US adults, 1999-2008. JAMA 2010;303(3):235e41. weathering and risk of preterm delivery. Am J Public Health 2009;99(10):1864e71. [4] Geronimus AT, Bound J, Waidmann TA, Hillemeier MM, Burns PB. Excess [36] Osypuk TL, Acevedo-Garcia D. Are racial disparities in preterm birth larger in mortality among blacks and whites in the United States. N Engl J Med hypersegregated areas? Am J Epidemiol 2008;167(11):1295e304. 1996;335(21):1552e8. [37] Hibbs S, Rankin KM, David RJ, Collins JW. The relation of neighborhood in- [5] Levine RS, Foster JE, Fullilove RE, Fullilove MT, Briggs NC, Hull PC, et al. Black- come to the age-related patterns of preterm birth among white and African- white inequalities in mortality and life expectancy, 1933-1999: implications American women: the effect of cigarette smoking. Matern Child Health J for healthy people 2010. Public Health Rep 2001;116(5):474e83. 2016;20(7):1432e40. [6] Wong MD, Shapiro MF, Boscardin WJ, Ettner SL. Contribution of major dis- [38] Sheeder J, Lezottte D, Stevens-Simon C. Maternal age and the size of white, eases to disparities in mortality. N Engl J Med 2002;347(20):1585e92. black, hispanic, and mixed infants. J Pediatr Adolesc Gynecol 2006;19(6):385e9. [7] Carter-Pokras O, Baquet C. What is a “health disparity”? Public Health Rep [39] Thorpe RJ, Fesahazion RG, Wilder T, Rooks RN, Bowie JV, Bell CN, et al. 2002;117(5):426e34. Accelerated health declines among African Americans in the USA. J Urban [8] Williams DR, Collins C. US socioeconomic and racial differences in health: Health 2016;93(5):808e19. patterns and explanations. Annu Rev Sociol 1995;21(1):349e86. [40] Geronimus AT, Bound J, Keene D, Hicken M. Black-white differences in age [9] Williams DR. Miles to go before we sleep racial inequities in health. J Health trajectories of hypertension prevalence among adult women and men, 1999- Soc Behav 2012;53(3):279e95. 2002. Ethn Dis 2007;17(1):40e8. [10] Krieger N. Embodying inequality: epidemiologic perspectives. New York, NY: [41] Lin J, Kelley-Moore J. Intraindividual variability in late-life functional limita- Routledge; 2016. tions among white, black, and hispanic older adults: implications for the [11] Colen CG. Addressing racial disparities in health using life course perspectives. weathering hypothesis. Res Aging 2017;39(4):549e72. Du Bois Rev 2011;8(01):79e94. [42] Chinn JJ, Hummer RA. Racial disparities in functional limitations Among [12] Adler NE, Stewart J. Health disparities across the lifespan: meaning, methods, Hispanic women in the United States. Res Aging 2016;38(3):399e423. and mechanisms. Ann N Y Acad Sci 2010;1186(1):5e23. [43] Spence NJ, Eberstein IW. Age at first birth, parity, and post-reproductive [13] Geronimus AT. Understanding and eliminating racial inequalities in women's mortality among white and black women in the US, 1982e2002. Soc Sci health in the United States: the role of the weathering conceptual framework. Med 2009;68(9):1625e32. J Am Med Women's Assoc (1972) 2000;56(4):133e6. 149-150. [44] Spence NJ. The long-term consequences of childbearing physical and psy- [14] Geronimus AT. The weathering hypothesis and the health of African- chological well-being of mothers in later life. Res Aging 2008;30(6):722e51. American women and infants: evidence and speculations. Ethn Dis [45] Wang S-C, Lee M-C. Effects of age, ethnicity and health behaviours on the prev- 1992;2(3):207e21. alence of adverse birth outcomes in Taiwan. J Biosoc Sci 2012;44(5):513e24. [15] Geronimus AT. Black/white differences in the relationship of maternal age to [46] Collins JW, Rankin KM, Hedstrom AB. Exploring weathering: the relation of birthweight: a population-based test of the weathering hypothesis. Soc Sci age to low birth weight among first generation and established United States- Med 1996;42(4):589e97. Born Mexican-American women. Matern Child Health J 2012;16(5):967e72. 18 A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18

[47] Bird CE, Seeman T, Escarce JJ, Basurto-Davila R, Finch BK, Dubowitz T, et al. [53] Howard JT, Sparks PJ. The effects of allostatic load on racial/ethnic mortality Neighbourhood socioeconomic status and biological ‘wear and tear’in a na- differences in the United States. Popul Res Policy Rev 2016;35(4):421e43. tionally representative sample of US adults. J Epidemiol Commun Health [54] Wallace ME, Harville EW. Allostatic load and birth outcomes among white and 2010;64(10):860e5. black women in New Orleans. Matern Child Health J 2013;17(6):1025e9. [48] Chyu L, Upchurch DM. Racial and ethnic patterns of allostatic load among adult [55] Geronimus AT, Hicken MT, Pearson JA, Seashols SJ, Brown KL, Cruz TD. Do US women in the United States: findings from the National Health and Nutrition black women experience stress-related accelerated biological aging?: A novel Examination Survey 1999e2004. J women's Health 2011;20(4):575e83. theory and first population-based test of black-white differences in telomere [49] Kaestner R, Pearson JA, Keene D, Geronimus AT. Stress, allostatic load, and length. Hum Nat 2010;21(1):19e38. health of Mexican immigrants. Soc Sci Q 2009;90(5):1089e111. [56] Geronimus AT, Pearson JA, Linnenbringer E, Schulz AJ, Reyes AG, Epel ES, et al. [50] Peek MK, Cutchin MP, Salinas JJ, Sheffield KM, Eschbach K, Stowe RP, et al. Race-ethnicity, poverty, urban stressors, and telomere length in a Detroit Allostatic load among non-hispanic whites, non-hispanic blacks, and people community-based sample. J Health Soc Behav 2015;56(2):199e224. of Mexican origin: effects of ethnicity, nativity, and acculturation. Am J Public [57] Strutz KL, Hogan VK, Siega-Riz AM, Suchindran CM, Halpern CT, Hussey JM. Health 2010;100(5):940e6. Preconception stress, birth weight, and birth weight disparities among US [51] Geronimus AT, Hicken M, Keene D, Bound J. “Weathering” and age patterns of women. Am J Public Health 2014;104(8):e125e32. allostatic load scores among blacks and whites in the United States. Am J [58] Das A. How does race get “under the skin”?: inflammation, weathering, and Public Health 2006;96(5):826e33. metabolic problems in late life. Soc Sci Med 2013;77:75e83. [52] Borrell LN, Dallo FJ, Nguyen N. Racial/ethnic disparities in all-cause mortality [59] Simons RL, Lei MK, Beach SR, Philibert RA, Cutrona CE, Gibbons FX, et al. in US adults: the effect of allostatic load. Public Health Rep 2010;125(6): Economic hardship and biological weathering: the epigenetics of aging in a US 810e6. sample of black women. Soc Sci Med 2016;150:192e200. A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18 18.e1

Appendix

Table A1 Sample form for data extraction (applied to each study that met the inclusion criteria in the review)

Data Data description Data categories

Publication characteristics

Author name Name of ﬁrst author Write in

Publication year Year the article was published Write in

Publication citation Full citation of the article Write in

Study characteristics

Geographic location Geographic location where the study was conducted 1 ¼ United States (specify state) 2 ¼ Other (specify)

Data source Source of the data Write in

Sample size Size of the study sample Write in

Study design Study design 1 ¼ Cross-sectional 2 ¼ Cohort 3 ¼ Case-comparison 4 ¼ Other (specify) 5 ¼ None speciﬁed

Physical health outcome Physical health outcome under study 1 ¼ Birth outcomes (specify type) 2 ¼ Cardiovascular (specify type) 3 ¼ Functional limitations (specify type) 4 ¼ Mortality (specify type) 5 ¼ Other (specify)

Participant characteristics

Age Age range of study participants Write in

Gender Gender of the study participants 1 ¼ Women 2 ¼ Men

Racial and/or ethnic group Race and/or ethnicity of study participants 1 ¼ African American or Black (specify) 2 ¼ Hispanic or Latinx (specify) 3 ¼ Asian (specify) 4 ¼ White (specify) 5 ¼ Other (specify)

Nativity Nativity status of participants 1 ¼ Native born (specify) 2 ¼ Foreign born (specify)

Weathering hypothesis characteristics

Type of test The test used to examine the weathering hypothesis 1 ¼ Age patterns 2 ¼ Modiﬁers of age distributions (specify) 3 ¼ Biological/physiological stress mechanisms (specify) 4 ¼ Other (specify)

Population comparisons Population comparisons used to test the weathering 1 ¼ African American or Black versus White (specify) hypothesis 2 ¼ Hispanic/Latinx versus White (specify) 3 ¼ Asian versus White (specify) 4 ¼ Other (specify)

Evidence Evidence regarding support for the weathering 1 ¼ Support (specify population, outcome, justification) hypothesis 2 ¼ No support (specify population, outcome, justification) 3 ¼ Partial support (specify population, outcome, justification) (continued on next page) 18.e2 A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18

Table A1 (continued )

Data Data description Data categories

Risk of bias characteristics

Participant selection Methods to select/recruit participants (e.g., medical Write in records)

Participant retention Location where participants were selected (e.g., Write in hospital)

Measurement of exposure List of exposures and methods for measuring exposures Write in

Measurement of outcome List of outcomes and methods for measuring outcomes Write in

Confounding variables List of confounding variables and methods for Write in measuring confounding variables

Statistical methods Data analyses description and power/sample size Write in assessment

Study limitations Limitations of the study Write in A.T. Forde et al. / Annals of Epidemiology 33 (2019) 1e18 18.e3

Table A2 Adapted version of the Quality in Prognostic Tool (adapted for this systematic review)

Domains Basis for judgment Ratings

Study participation a Adequate participation in the study by eligible persons High risk of bias: The relationship between the exposure and b Description of the source population or population of interest outcome is very likely to be different for participants and eligible nonparticipants c Description of the baseline study sample

d Adequate description of the sampling frame and recruitment Low risk of bias: The relationship between the exposure and e Adequate description of the period and place of recruitment outcome may be or is unlikely to be different for f Adequate description of inclusion and exclusion criteria participants and eligible nonparticipants

Study attrition a Adequate response rate for study participants High risk of bias: The relationship between the exposure and b Description of attempts to collect information on participants outcome is very likely to be different for completing and who dropped out noncompleting participants c Reasons for loss to follow-up are provided Low risk of bias: The relationship between the exposure and d Adequate description of participants lost to follow-up outcome may be or is unlikely to be different for completing e There are no important differences between participants who and noncompleting participants completed the study and those who did not

Exposure measurement a A clear deﬁnition or description of the exposure is provided High risk of bias: The measurement of the exposure is very b Method of exposure measurement is adequately valid and likely to be different for different levels of the outcome of reliable interest c Continuous variables are reported or appropriate cut points Low risk of bias: The measurement of the exposure may be are used or is unlikely to be different for different levels of the d The method and setting of measurement of exposure is the outcome of interest same for all study participants e Adequate proportion of the study sample has complete data for the exposure f Appropriate methods of imputation are used for missing exposure data

Outcome measurement a A clear deﬁnition of the outcome is provided High risk of bias: The measurement of the outcome is very b Method of outcome measurement used is adequately valid likely to be different related to the baseline level of the and reliable exposure c The method and setting of outcome measurement is the same Low risk of bias: The measurement of the outcome may be or for all study participants is unlikely to be different related to the baseline level of the exposure

Study confounding a All important confounders are measured High risk of bias: The observed effect of the exposure on the b Clear deﬁnitions of the important confounders measured are outcome is very likely to be distorted by another factor provided related to the exposure and outcome c Measurement of all important confounders is adequately Low risk of bias: The observed effect of the exposure on the valid and reliable outcome may be or is unlikely to be distorted by another d The method and setting of confounding measurement are the factor related to the exposure and outcome same for all study participants e Appropriate methods are used if imputation is used for missing confounder data f Important potential confounders are accounted for in the study design g Important potential confounders are accounted for in the analysis

Statistical analysis and reporting a Sufﬁcient presentation of data to assess the adequacy of the High risk of bias: The reported results are very likely to be analytic strategy spurious or biased related to analysis or reporting b Strategy for model building is appropriate and is based on a conceptual framework or model Low risk of bias: The reported results may be or are unlikely to be spurious or biased related to analysis or reporting c The selected statistical model is adequate for the design of the study d There is no selective reporting of results

Source for the original QUIPS tool: Hayden JA, van der Windt DA, Cartwright JL, Cot^ e P, Bombardier C. Assessing bias in studies of prognostic factors. Ann Intern Med 2013; 158 (4):280-6.