Longitudinal associations of neighborhood socioeconomic status with cardiovascular risk factors: a 46-year follow-up study

Abstract

Objectives.

Neighborhood characteristics are increasingly recognized as important determinants of cardiovascular disease (CVD) risk. However, longitudinal studies on the health impacts of neighborhood characteristics are rare. We sought to investigate whether neighborhood socioeconomic status (NSES) during birth, childhood and adulthood is associated with CVD risk factors in adulthood.

Methods.

Using longitudinal data from the New England Family Study (n=671) with 46-years of follow-up, participants’ home addresses were geocoded at birth (mean age=1.6 months), childhood (mean age=7.1 years), and adulthood (mean age=44.4 years) across Massachusetts and Rhode Island in the US from 1961 to 2007. We used multilevel models to evaluate associations of NSES across the life-course with systolic blood pressure, diastolic blood pressure and body mass index (BMI) in adulthood, adjusting for age, sex, race/ethnicity, mother’s race, individual SES, and parental SES.

Results.

In fully adjusted models, one standard deviation higher NSES at birth was associated with a 1.9 mmHg lower SBP (95% CI: −3.8, −0.1) and 1.3 mmHg lower DBP (95%CI:−2.6,−0.03) in adulthood; while one standard deviation of higher NSES at adulthood was associated with 0.87 kg/m² lower BMI (95%CI: −1.7, −0.1).

Conclusions.

We found that living in a socioeconomically disadvantaged neighborhood early in life and in adulthood was associated with blood pressure and BMI, respectively, two established risk factors for CVD. Our findings support a longitudinal association between exposure to socioeconomically disadvantaged neighborhoods in early life and CVD risk factors in adulthood.

Despite significant declines in mortality in the U.S., cardiovascular disease (CVD) remains the leading cause of death with considerable burden on both health and costs. Hypertension and obesity, two major risk factors for adult CVD, are highly prevalent in the U.S.; 70 million (29%) U.S. adults have hypertension ^1–3 and more than a third are classified as obese.⁴

Many studies have documented inverse gradients in the incidence, morbidity and mortality of CVD across the spectrum of markers of socioeconomic status.⁵ The additional contribution of the socioeconomic conditions of the neighborhood of residence to the risk of CVD, independent of individual socioeconomic status, is increasingly a concern. For example, comparisons of obesity prevalence using geographic criteria indicate that residing in low income areas is associated with higher risks of obesity ⁶ and coronary heart disease,⁷ even after adjusting for individual-level markers of socioeconomic status. Neighborhood characteristics may play an important role in the development of CVD risk factors through a range of both physical and social mechanisms. Neighborhood socioeconomic status (NSES) may affect health by limiting opportunities for education and employment.⁸ Furthermore, deprived neighborhoods are likely to experience financial disinvestment, low social cohesion, and poor collective efficacy (ability of residents to improve their neighborhoods), leading to deteriorating buildings, dirty streets, crime, and lower walkability.⁹ These conditions may contribute to lower access to and quality of resources, increased levels of stress, and unhealthy behaviors—such as limited physical activity, poor diet, and smoking— all of which are risk factors for CVD.^10–12

Most prior studies have investigated cross-sectional associations between neighborhood deprivation and CVD risk factors, and thus do not provide direct evidence regarding the long-term health impacts of NSES, nor allow exploration of stages across the life-course when individuals may be particularly susceptible. During the past two decades, pathological data showed that atherosclerosis begins in childhood and predicts adulthood CVD risk.¹³ Furthermore, early childhood is a sensitive period in the life-course during which inputs received can set health-related behaviors that may have long lasting impacts.¹⁴ Thus long-term neighborhood exposures may be more relevant to investigate than single-point-in-time exposures.¹⁵ Very few studies have documented longitudinal exposure to NSES in association to health outcomes (including CVD) later in life.^15–19 We use two theoretical perspectives to examine the longitudinal neighborhood exposure in association to CVD risk factors later in life. First, the sensitive periods model holds that negative health exposures have greater effects when they occur during developmentally vulnerable stages in life and affect later-life outcomes irrespective of future exposures.²⁰ This model has been used to explain associations between early-life disadvantage and future outcomes, including adult life expectancy.²¹ Under this model, children in deprived neighborhoods are expected to have a higher risk of CVD in adulthood versus those from non-deprived neighborhoods, irrespective of residential circumstances in adulthood. Second, the cumulative disadvantage theory²² states that detrimental exposures at multiple life-stages operate jointly to affect health. Under this model, CVD risk is expected to be worst among those who consistently live in deprived neighborhoods, best among those who never live in deprived neighborhoods, and somewhere in between for those who fluctutated (enter or exit neighborhood poverty).¹⁶ The ability to track neighborhood exposures longitudinally¹⁷ and the development of strategies that allow linkage of ongoing cohorts to historical spatial data is critical to test these theoretical perspectives and advance this area of research.²³ In the present study, we linked historical geographic data to a 48-year-long cohort to test sensitive period and cumulative disadvantage models of the association between NSES and CVD risk factors.

METHODS

Data

We used data from the New England Family Study (NEFS), a longitudinal investigation with up to 48 years of follow-up of the adult offspring of pregnant women enrolled between 1959 and 1966 in the Collaborative Perinatal Project (CPP). The NEFS was initiated to locate and interview adult offspring in the Providence, Rhode Island, and Boston, Massachusetts CPP sites.²⁴ Analyses for this project were achieved by merging data from two of the follow-up studies that comprise NEFS: Longitudinal Effect on Aging Perinatal Project (LEAP) and Pathways Linking Education and Health in Middle Adulthood Project (Edhealth). Between 2010 and 2011, 400 CPP adult offspring (mean age= 46.1) participated in a study of the early life origins of ageing in mid-life; between 2005 and 2007, 618 CPP adult offspring (mean age=42.5) participated in Edhealth, its main purpose was to assess pathways linking education and health.²⁵ Collectively, this yielded an analytical dataset of 931 participants, including 113 sibling sets.²⁶

Residential addresses for the participants were collected at 12 different occasions from birth (mean age=1.60 months, SD=4.5 months, between January 1961 and June 1967) through childhood (mean age=7.08 years, SD=0.6 years, between January 1970 and May 1973) (see Supplemental Table 1 for details), and then again at adulthood (mean age=44.41 years, SD=2.9 years, around 2000s). For the purposes of this analysis, childhood was defined as ages 7-8 years which was the time of the last child assessment. Addresses were subsequently geocoded using the Brown University Geocoding Service (96% of the total number of addresses),²⁷ and World Geocoding Services (3% of the total number of addresses, which did not report a street number).²⁸ Of the 931 participants assessed in adulthood, 805 had address information available at birth, and 724 at childhood (Supplemental Table 1). A total of 692 participants had address information available at all time-points from birth, childhood and adulthood within MA or RI. Of these, the final analytic sample included n=671 who also had complete data on systolic blood pressure (SBP), diastolic blood pressure (DBP) and body mass index (BMI) in adulthood respectively (Supplemental Figure 1). All spatial analyses were performed using ArcGIS, Version 10.4.1 (Redlands, CA).

Measures

NEFS cohort data were linked to historical geographic data. The primary exposure was NSES measured at the census-tract level (CT) throughout the life-course. Spatial data on NSES were drawn from the National Historical Geographic Information System,²⁹ and the Bureau of the Census; the final variables were created by the staff of the American Communities Project at Brown University.³⁰ Defining neighborhoods based on census-tracts is commonplace among studies in the US.³¹ Because of changing census-tracts borders over time, CT boundaries from 1970 were homogeneized to 1960’s boundaries, following the methodology used by MApUSA, a project gauging census data at the tract level from 1940 to 2010.³² Homogeneization of the 2000 census tracts was not needed since less than 1% (n=5) participants stayed in the same household from birth to adulthood, which aligns well with previous research showing that residential mobility peaks in early adulthood.¹⁶ Census data were assigned to subjects by waves according to their CT of residence and the date of their study examination, selecting the Census closest to the examination date.

At each time point, NSES was measured as a standardized score (mean=0, standard deviation=1) derived from the following census variables: % of residents with less than a high school education, % of residents unemployed, and median household income.^33,34 Household income from the 1960s and 1970s was adjusted for inflation rates to 2000 dollars.³⁵ Income, education and unemployment are the most standard variables, either separately or jointly, used to characterize neighbourhood disadvantage and deprivation in other studies.^7,33,36 Lower standardized scores indicated lower NSES. The NSES score was also dichotomized below (low NSES) and above zero (high NSES)³⁷ to create an indicator of NSES mobility as follows: (1) participants who remained categorized as low NSES throughout the life-course, (2) participants who remained categorized as high NSES throughout the life-course, and (3) everyone else (those who fluctuated from low to high, high to low or some combination).

Systolic blood pressure (SBP), diastolic blood pressure (DBP) and body mass index (BMI) in adulthood were the primary outcome variables. Five SBP and DBP measurements were attained over one-minute intervals in the right arm at heart level in participants seated, after resting for 5 minutes, using automated blood pressure monitors in EdHealth SBP and DBP were computed by taking the average of the lowest three SBP or DBP readings, excluding the first measurement.³⁸ In LEAP, three blood pressure measurements were assessed by certified research nurses using mercury sphygmomanotmetes in seated participants resting 5 minutes prior to assessment consistent with American Heart Association guidelines. The mean of the second and third blood pressure readings were used. Height and weight were measured with participants wearing light clothing without shoes using calibrated stadiometers and weighing scales. BMI was calculated as weight (kg) divided by height (m) squared. Covariates included age at adult interview, educational attainment (less than high-school, high-school, more than high-school), mother’s race, mother’s and father’s education measured at birth during the CPP (less than high-school, high-school, more than high-school), and participants’ sex and race/ethnicity at adulthood (White, African-American, Hispanic, other). Please refer to supplemental Figure 2 for more detail on covariate selection.

Statistical Analysis

We fit separate sets of models to test the sensitive period and the cumulative risk hypotheses. To test whether NSES at birth was a sensitive period relative to NSES at childhood and adulthood, we estimated 3 models and compared the coefficients for NSES. Model 1 estimated the impact of NSES at birth without adjustment for NSES at childhood or NSES at adulthood. Model 2 estimated the impact of NSES in childhood, adjusted for NSES at birth. Model 3 estimated the impact of NSES in adulthood, adjusted for NSES at birth and childhood.

The second set of models evaluated the cumulative disadvantage model, where the main exposure was NSES mobility throughout the life-course. To test whether persistent exposure to low NSES versus high NSES or fluctuating NSES throughout the life-course was associated with a higher blood pressure and BMI, we used the indicator of NSES mobility as our main exposure in the models for all three outcomes.

We used multilevel models to account for the fact that observations were hierarchically nested, such that members of the lower level (i.e., level one) are nested in one and only one entity at the higher levels (i.e., levels two and three). We used three-level models to account for individuals (level 1) nested within families (level 2) and within neighborhoods (level 3) by including random intercepts for each family and for each census tract. We controlled for the following covariates in all analyses: age, sex, race/ethnicity, mother’s race, individual SES, and parental SES. Two-sided confidence intervals were reported. Our main results are based on a complete-case analysis. However, we also conducted multiple imputation as a sensitivity analysis (Supplementary Table 1) to evaluate the effect of missing data on our results. For the Multiple Imputation procedure, we used Multiple Imputation by Chained Equations to incorporate the multi-level structure of the data, the “mice” package in R software with 10 imputations. Finally, to account for repeated measures with changing neighborhood membership over time, we used a cross-classified multilevel model, with time as a level 1, and individual and neighborhood were separate cross-classified levels for level 2. Data management was performed in SAS version 9.4 and statistical analyses were performed using R version 3.4.1 and MLWin 2.34.

RESULTS

The analysis sample included 671 offspring, of which 59% were female. The mean age at adult follow-up was 44.2 years. Study participants were 76% White, 18% African-American, 2% Hispanic, and 4% were categorized as other race/ethnicity. About 12% of the sample had less than high-school education, and paternal education level was less than high-school for more than half of the sample (Table 1). Maternal race was White for 77% of the sample. The average values for the outcomes were: BMI 29.9 kg/m², SBP 117.5 mmHg and DBP 76.4 mmHg (Table 1). Demographic characteristics of the participants included in the analytical sample (N=671) compared to the full sample (N=931) were not statistically significantly different. Participants lived in 256 different census tracts at birth, 361 at childhood, and 425 at adulthood. On average, there were 2.6 participants per census tract at birth, 1.8 at childhood, and 1.6 at adulthood. Descriptive statistics of the NSES variables demonstrated that household income increased across time, while percentage of unemployment and percentage with less than high-school education decreased (Table 2). NSES characteristics of the NEFS participants were comparable to those of all MA and RI. The correlation between NSES at birth and at childhood was 0.70, which decreased to 0.41 between birth and adulthood; and the correlation between NSES at childhood and adulthood was 0.44. Indicators of neighborhood socio-economic mobility showed that 174 (26%) participants lived in neighborhoods with low NSES at all time-points, 286 (43%) fluctuated in NSES (i.e. went from low to high, high to low, or some other combination of NSES across the life-course), and the remaining 211 (31%) lived in neighborhoods with high NSES at all time points. From the 286 that fluctuated in NSES, some of the trajectories seen were: 32% were in low NSES at birth and childhood but high NSES at adulthood, 24% were in high NSES at birth and childhood but low NSES at adulthood, and 13% were in low NSES at birth, high NSES at childhood, and low NSES again at adulthood. The distribution of co variates and outcomes across NSES throughout the life-course showed that participants in neighborhoods with high NSES at birth, childhood and adulthood had lower BMI, SBP and DBP (Supplemental Tables 2–3).

Table 1.

Descriptive Characteristics of New England Family Study Participants (N=671).

Variable Description	N	%
Female (%)	397	59.2
Race (%)
White	507	75.6
African-American	123	18.3
Hispanic	16	2.4
Other	25	3.7
Education level (%)
Less than High-school	82	12.2
High-school	300	44.7
More than High-school	289	43.1
Mother’s Race (%)		0.0
White	518	77.2
Non-White	153	22.8
Mother’s Education level (%)
Less than High-school	337	50.2
High-school	233	34.7
More than High-school	101	15.1
Father’s Education level (%)		0.0
Less than High-school	368	54.8
High-school	204	30.4
More than High-school	99	14.8
	Mean	SD
Age, y	44.2	2.9
Body Mass Index, kg/m2	29.9	7.7
Systolic Blood Pressure, mmHg	117.5	16.2
Diastolic Blood Pressure, mmHg	76.4	11.0

Table 2.

Census Tract Socioeconomic Characteristics of New England Family Study Participants, N=671

	Birth		Childhood		Adulthood
	Mean	SD	Mean	SD	Mean	SD
Median Household Income *	27,763	8,389	35,911	11,210	47,309	21,276
% Unemployed	6.1	2.7	4.2	2.2	4.1	3.5
% With High-school degree or less	91.2	14.0	88.6	11.8	49.0	16.1
NSES score	−0.002	0.7	0.01	0.8	0.01	0.9

^*Please note that income from 1960s and 1970s were adjusted for inflation rates to 2000, and thus are reflective of 2000 dollars.

In models evaluating sensitive periods for SBP, one standard deviation higher NSES at birth was associated with 1.9 (95% CI: −3.7, −0.1) mmHg lower SBP and 1.3 (95% CI: −2.6,−0.03) mmHg lower DBP, adjusting for for age, gender, race, and individual-level and parental socioeconomic status (Table 3). NSES in childhood and adulthood were less strongly associated with SBP and DBP and were not statistically significant. In contrast, NSES at birth and childhood were not associated with BMI in adulthood. However, one standard deviation higher NSES during adulthood was associated with 0.87 (95% CI: −1.7, −0.1) kg/m² lower BMI. Models using multiple imputation showed similar directions of association, with smaller estimates for SBP and DBP, and similar estimates for BMI (Supplemental Table 4).

Table 3.

Three-level Models Demonstrating Adjusted Associations of Neighborhood Socioeconomic Status (NSES) and Adult Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP) and Body Mass Index (BMI), Stratified by Life-Course Stage that NSES was Assessed (N=671).

	SBP		DBP		BMI
	Estimate	95% Confidence Limits	Estimate	95% Confidence Limits	Estimate	95% Confidence Limits
NSES z-score at Birth1	−1.92	−3.77, −0.08 *	−1.31	−2.59, −0.03*	−0.41	−1.35, 0.52
NSES z-score at Childhood2	−1.28	−3.25, 0.69	−0.64	−2.00, 0.73	−0.31	−1.31, 0.68
NSES z-score at Adulthood3	−1.48	−3.04, 0.07	−0.96	−2.04, 0.12	−0.87	−1.66, −0.07*

¹Adjusted for age, gender, race, individual and parental SES

²Adjusted for age, gender, race, individual SES, parental SES, and birth NSES

³Adjusted for age, gender, race, individual SES, parental SES, birth and childhood NSES

^*P <0.05

We next compared levels of adult SBP, DBP, and BMI among participants that lived in neighborhoods with low versus high NSES at each time point (cumulative disadvantage model, Table 4). Compared to participants that lived in low NSES neighborhoods at each time point, participants who had fluctuating NSES over the life-course had a 3.4 (95% CI: −6.4, −0.3) mmHg lower SBP in adulthood, whereas those that stayed in high SES had no statistically significant difference. No other statistically significant associations were observed.

Table 4.

Three-level Models for Adjusted Associations of Neighborhood Socioeconomic Status (NSES) Mobility and Adult Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP) and Body Mass Index (BMI) (N=671).

	SBP		DBP		BMI
Models	Estimate	95% Confidence Limits	Estimate	95% Confidence Limits	Estimate	95% Confidence Limits
NSES Mobility (ref: stayed in low NSES)
Fluctuated	−3.37	−6.40, −0.35*	−1.35	−3.46, 0.76	−0.33	−1.86, 1.21
Stayed in high NSES	−2.51	−6.12, 1.10	−1.27	−3.79, 1.25	−1.11	−2.94, 0.72

^*P <0.05

DISCUSSION

Data from a longitudinal cohort of 671 participants in the New England Family Study were used to investigate associations between NSES across the life-course and CVD risk factors. The study incorporated objective measurements of neighborhood characteristics, had up to 48-years follow-up from birth through adulthood, and used standardized assessments of blood pressure and body mass index. Analyses showed that neighborhood socioeconomic disadvantage during certain periods of the life-course was related to higher prevalence of CVD risk factors. Specifically, we found that higher NSES at birth was associated with lower blood pressure in adulthood, while higher NSES in adulthood was associated with lower contemporaneous BMI. This research adds to prior literature by directly testing the sensitive period and the cumulative risk models for NSES across the life-course, and highlights the potential role of NSES at specific periods of the life-course on adult cardiovascular health.

Prior evidence about associations between NSES across the life-course and CVD risk factors is limited, as most studies have been cross-sectional in design ^39–42. The first set of models of the present study tested the sensitive period model. Our results suggested that each of the 3 measures of neighborhood deprivation contributed to an increased risk of hypertension or obesity later in life. We hypothesized childhood would be a sensitive period for exposure to neighborhood socioeconomic disadvantage and the study’s 3 CVD risk factors, as this is a time when health behaviors such as food taste preference and physical activity norms may be set ⁴³, but found that sensitive periods differed across the study’s outcomes. For instance, we found evidence that NSES at birth rather than in childhood or adulthood had a stronger association with SBP and DBP; while NSES in adulthood had a stronger association with BMI than NSES at birth and childhood. These differences in BP and BMI by life-course stage are unique in the literature. It will be important to attempt to replicate these findings in other studies to evaluate the relative importance of these unique life-course stages, and likely mechanisms.

Observations in animals show that the environment (ex: lead and air pollution, neighborhood socioeconomic status) during development can permanently change the body’s structure and function as well as its responses to environmental influences encountered in later life, which may affect the regulation of blood pressure and metabolism ⁴⁴. The current study suggests that exposure to low NSES during the first year of life has stronger associations with blood pressure than exposure to low NSES at later time points. Low NSES at birth can affect the access to healthy food during this crucial stage of development which in turn has consequences on the regulation of blood pressure ⁴⁵. Studies investigating child growth and socioeconomic circumstances suggest that the pathogenesis of CVD may be a product of branching paths of development that are triggered by the socioeconomic environment ^46,47. A recent analysis of the same cohort showed that infancy may be a sensitive period for exposure to family socioeconomic disadvantage, as exposure in the earliest years of life confers a larger risk for overall and central adiposity in mid-adulthood than exposure during childhood²⁵. On the other hand, adulthood NSES showed a stronger association with BMI. Future adequately powered mediation analysis may be able to evaluate mechanisms through which NSES is associated with BMI and BP. Perhaps specific mechanisms, such as early life adversity associated with low childhood NSES, trigger stress reactivity pathways that would then influence adult blood pressure through it’s strong sympathetic nervous system responsivity, and have a weaker influence on adult BMI which is less influenced by the sympathetic nervous system. Research clarifying the importance of NSES at specific life-course stages is needed to develop effective policy strategies.

The second set of models tested the cumulative disadvantage model, which is particularly important for diseases that develop over a long period of time, as is the case of CVD. Our results show no association between persistent exposure to low NSES and CVD risk factors. While effect sizes were in favorable directions for lower levels of blood pressure and BMI in those exposed to high NSES across the life-course, analyses may have either been underpowered to show associations, or the magnitude of the association may be small. However, we did find that participants who had fluctuating NSES over the life-course had a 3.4 (95% CI: −6.4, −0.3) mmHg lower SBP in adulthood. This result might have been driven by upward mobility, participants who went from low to high NSES; nevertheless we did not have sufficient statistical power to test this difference.

There were several strengths to these analyses. Specifically, our study is unique in using longitudinal data with a 48-year-follow-up to estimate the effects of NSES on CVD risk factors. Second, we are able to identify critical stages in the life-course where NSES may slow development of CVD in adults. One important obstacle for causal inference in neighborhood research is neighborhood self-selection due to the fact that individuals may personally select their place of residence based on their predisposition to certain health behaviors (e.g. people who are more inclined to be physically active may choose to live in areas with better physical activity resources) ^23,48. A major strength in this analysis, is that neighborhood’s features were assessed at birth and childhood, when participants were likely not able to select their place of residence. Our study provides support that the neighborhood socio-economic environment influences cardiovascular health. Differences in sensitive periods further suggest that neighborhood-level interventions may benefit from being timely-targeted to individual outcomes. The statistical analysis used for this analysis allowed us to evaluate between- and within-unit variability at more than one level. Multilevel models still work at least as well as classical regression even when sample sizes are small within families or census tracts, since the key concern is the estimation of variance parameters.⁴⁹ Thus we are confident that the data still provide partial information that allowed estimation of the coefficients and variance parameters of the individual- and group-level regressions.

Limitations

Finally, there are limitations to our research. First, we have missing data due to lack of addresses available and therefore neighborhood data that may result in selection bias. However, we used MICE and sampled imputed values from the posterior predictive distribution. This imputation model accounts for the clusters in our data and for the process that created the missing data. The results from the sensitivity analysisis using MICE showed a similar trend in the relative importance of early and later-life exposures by health outcome, although the association was weakened. In addition, our analysis includes only participants who resided within MA or RI as adults, which can reduce generalizability to other populations. Further analysis is needed to establish whether these findings generalize to individuals living elsewhere. This was a strategic decision made, as we were interested in using different indicators of neighborhood deprivation, and while the sources of data for neighborhood indicators were comparable for MA and RI, this was unfortunately not the case for other states. Second, we might be missing relevant socioeconomic variables in our assessment of NSES, and thus the results can be biased towards the null, especially since the magnitude of associations observed were small. We looked into adding other socioeconomic variables, however unfortunately not all variables were measured consistently throughout time since 1960 and thus we were only able to include the 3 variables in the NSES score. Third, while the inclusion of NSES at other time points in this study is a step forward in neighborhood research, the absence of inclusion of time-varying information for the outcomes or other study covariates makes it difficult to deal with neighborhood selection issues, so our results could still be biased. Fourth, the lack of antihypertensive medication data biases results towards the null, since we are underestimating SBP and DBP for participants taking antihypertensive medications. Fifth, since census tracts boundaries change across time, the strength of the associations found at different life periods may potentially be due to changes in census boundaries over time. We used a cross-classified structure to account for changing neighborhoods over time and results were similar. In addition, exposure to NSES at different time periods may not hold the same meaning, and thus the intrepretation of the associations at different time periods might differ. Finally, we cannot rule out residual confounding.

Conclusion

In conclusion, analyses demonstrated that NSES at birth was associated with SBP and DBP, whereas NSES during adulthood was associated with BMI. This provides evidence suggesting that exposure to NSES at specific stages during the life-course may be important for understanding and potentially predicting CVD risk.

Research highlights.

• Exposure to neighborhood poverty has long-lasting (46-year) relations to health.

• Sensitive periods of exposure differ among cardiovascular outcomes.

• Higher neighborhood poverty at birth positively affects adult blood pressure.

• Higher neighborhood poverty at adulthood positively affects adult obesity.