[Obesity Discussion]

Life-Course Socioeconomic Position and Obesity in African American Women: The Pitt County Study

Objectives. We studied obesity in African American women in relationship to their socioeconomic position (SEP) in childhood and adulthood.

Methods. On the basis of parents' occupation, we classified 679 women in the Pitt County (North Carolina) Study into low and high childhood SEP. Women's education, occupation, employment status, and home ownership were used to classify them into low and high adulthood SEP. Four life-course SEP categories resulted: low childhood/low adulthood, low childhood/high adulthood, high childhood/low adulthood, and high childhood/high adulthood.

Results. The odds of obesity were twice as high among women from low versus high childhood SEP backgrounds, and 25% higher among women of low versus high adulthood SER Compared to that in women of high SEP in both childhood and adulthood, the odds of obesity doubled for low/low SEP women, were 55% higher for low/high SEP women, and were comparable for high/low SEP women.

Conclusions. Socioeconomic deprivation in childhood was a strong predictor of adulthood obesity in this community sample of African American women. Findings are consistent with both critical period and cumulative burden models of life-course socioeconomic deprivation and long-term risk for obesity in African American women. (Am J Public Health. 2006;96:554-560.)

Obesity is an increasingly prevalent condition in the United States, particularly among African American women, who currently have the highest prevalence of obesity of any US demographic group. According to the 1999-2002 National Health and Nutrition Examination Survey, 49% of African American women are obese, compared to 38.4% of Mexican American women and 30.7% of White women.(n1) The development of obesity is multifactorial, resulting from an interaction between an individual's genetic makeup and his/her health behaviors (e.g., diet, physical activity, smoking), with the latter being strongly patterned by one's access to both individual and neighborhood-level socioeconomic resources over the entire life course.(n2-n6)

Indeed, a life-course perspective on the problem of adulthood obesity is receiving increased attention from public health researchers in Europe,(n6-n9) the United States,(n10) and Latin America,(n11) with most studies indicating that low SEP in childhood, like low SEP in adulthood, is associated with increased risk for adulthood obesity. A life-course perspective on socioeconomic conditions and obesity would seem to be especially important in the case of African American women, a group known to be at increased risk for lifelong poverty(n12,n13) as well as for adulthood obesity.(n1,n14)

To date, however, inconsistent findings characterize the handful of studies investigating the association between childhood SEP and future risk of obesity in African American women. For example, in a study from the National Longitudinal Survey of Youth,(n15) researchers found that parental education was inversely associated with risk of adulthood obesity in African American women. Another study of African American and White adults in 4 US cities found that for White women, but not African American women, parental education was inversely associated with mean body mass index (BMI), independent of adulthood SEP.(n10) A Philadelphia-based study that followed a group of African Americans from birth to young adulthood likewise failed to observe an inverse association between maternal education and adult adiposity.(n16) Finally, although associations between intergenerational social mobility or the movement from one SEP level in childhood to another in adulthood have been studied in European populations,(n7,n8) we found no similar studies of African Americans. Given the growing public health interest in how socioeconomic deprivation over the life course might accelerate deterioration in health by early/middle adulthood and the limited research of this kind focusing specifically on the excess risk for obesity in African American women, additional studies are clearly needed.

Our aims were 2-fold: (1) examine the independent associations between childhood and adulthood SEP and risk for obesity in adulthood in a community probability sample of southern, African American women and (2) examine the degree to which the trajectories of obesity risk initiated in childhood are altered by differential changes in women's relative access to socioeconomic resources between childhood and adulthood.

METHODS
Study Participants
Data for this study came from the 2001 follow-up interviews of participants in the Pitt County Study, a community-based, prospective investigation of risk factors for hypertension and related disorders in African Americans who were aged 25 to 50 years in 1988, the baseline year. Because a major objective of the Pitt County Study was to investigate differential risk for hypertension between working-class and middle-class African Americans, individuals residing in middle-class neighborhoods were oversampled. The baseline sample, the sampling strategy, and the content of the baseline household interview are described elsewhere.(n17,n18)

Of the 2225 race- and age-eligible individuals, 1773 (661 men and 1112 women), or 80%, were interviewed in 1988. In 2001, the cohort was reinterviewed to obtain information on the individuals' social and economic resources from early childhood to the date of the interview. The goal was to link this information on socioeconomic resources to major cardiovascular disease risk factors, such as hypertension, obesity, and cigarette smoking, as recorded in 1988.

Interviews in 2001 were sought with all cohort members believed to be alive, noninstitutionalized, and residing within a 100-mile radius of Greenville, the county's principal city. Of the 1540 individuals (543 men and 997 women) meeting these criteria, 1221 (428 men and 793 women; 79%) were reinterviewed. Of these, 43 were excluded because of significant discrepancies in birth year (≥ 2 years) or height (≥ 2 inches) when comparing 1988 and 2001 values. These exclusions resulted in 1178 individuals (418 men and 760 women), or 77% of the 1540 targeted interviews. This report focuses on the women respondents.

Measurement of Obesity
In 1988, trained interviewers measured weight (in pounds) with a balanced scale after study participants removed their shoes and heavy clothing. Participants' height to the nearest inch was measured with a vertical ruler; BMI was computed as weight (in kilograms) divided by height (in meters), squared.(n19) In keeping with Centers for Disease Control and Prevention guidelines,(n20) the following BMI cutpoints were used to characterize the sample: obese (BMI≥30.0), overweight (BMI = 25.0-29.9), normal weight (BMI = 18.5-24.9), and underweight (BMI<18.5).

Measurement of Childhood SEP
Data to measure childhood SEP were obtained with the assistance of a computerized Event History Calendar (EHC). This methodology enhances recall of information stored years or decades in the past by using more easily remembered events (e.g., where one lived and with whom at specific points in time) to stimulate the recall of events less easily remembered.(n21-n23) Study participants were asked to provide a brief description of the main job held by their family's primary earner during their childhood years, which, in the current study, refers to the period between birth and 13 years of age. The primary earner could be the respondent's biological parent, grandparent, or a biologically unrelated person. Each job description was coded to fit 1 of 9 categories of the 1990 Census Occupational Classification: 1 = managerial and professional; 2 = technical, sales, and administrative support; 3 = protective services (including military); 4 = farm owners; 5 = precision production, craft, and repair; 6 = service occupations for private households; 7 = service occupations, except protective and households; 8 = operators, fabricators, assemblers, and laborers; and 9 = farm laborers. No code exists for "home-maker." In the case of 2 salaried working parents, the higher occupational rank, irrespective of gender, was used.

These 9 job categories were subsequently collapsed into 2 broad categories: skilled (codes 1-5) versus unskilled (codes 6-8) or farm laborer (code 9), and designated high and low childhood SEP, respectively. Childhood SEP could not be determined for 45 women because 27 were offspring of single mothers who had never worked outside the home (hence, homemakers) and 18 women had missing data for the primary earner variable.

Measurement of Adulthood SEP
Our prior work(n18,n24) indicated that education and occupation, taken alone or in combination, were weak predictors of hypertension-related outcomes in the Pitt County Study population. Therefore, in the current study we sought to minimize misclassification of respondents with respect to their "true" socioeconomic standing in the community by creating an index of adulthood socioeconomic position based on 4 variables that were collected in 1988.

The first variable, education, had 4 levels: less than high school, high school, some college, and college graduate. The second variable, occupation, was based on 9 Hollingshead job prestige(n25) scores: 1 = farm laborer/ menial service worker, 2 = unskilled worker, 3 = machine operator or semiskilled worker, 4 = skilled manual worker, 5 = clerical/sales worker, 6 = skilled technician/small business owner, 7 = manager/farm owner (> 150 acres), 8 = administrator/registered nurse, and 9 = higher executive/major professional. These 9 scores were subsequently collapsed into 2 broad occupational categories: "blue collar" for Hollingshead scores from 1 to 4 and "white collar" for Hollingshead scores from 5 to 9. The third variable, employment status, had 2 levels: employed versus not employed, as did the fourth variable, homeowner: yes or no. Household income was not collected in 1988; hence, employment status and home ownership provided some indirect information on respondents' differential access to income and wealth.

Scores for the adulthood SEP index were produced with the following algorithm: education (less than high school = 0, high-school graduate but less than college = 0.5, college graduate = 1.0); occupation (blue collar = 0, white collar = 1); currently employed (no = 0, yes = 1); and home owner (no = 0, yes = 1). Hence, the highest possible score on the adulthood SEP index was 4.0. To identify a subset of individuals who could be plausibly designated "socioeconomically advantaged" relative to other cohort members, those who scored 3.0 or higher on the adulthood SEP index were categorized as "high"; those who scored less than 3.0 were categorized as "low."

Measurement of Life-Course SEP
Life-course SEP was determined by combining information on childhood and adulthood SEP. Four nonoverlapping life-course SEP categories were created: low childhood/ low adulthood, low childhood/high adulthood, high childhood/low adulthood, and high childhood/high adulthood.

Covariates
Potential confounders, measured in 1988, included the following: age (years), marital status (currently/formerly/never married), alcohol consumption (abstainer/drinker), and cigarette smoker (yes/no). The EHC was used to collect information on a fifth potential confounder, food insecurity during childhood, was obtained in 2001. Individuals who answered "no" for any year during the first 13 years of life to the question, "Did you have enough food to eat?" were considered exposed to childhood food insecurity, theoretically a condition that could be correlated with both childhood SEP and body weight in adulthood. Data from 1988 on the following potential intervening variables were also included in the analyses: strenuous physical activity (≥ 3 times/week, ≥20 minutes per occasion, intense enough to breathe hard and perspire); daily vegetable consumption (low≤1 serving/day); and daily fruit consumption (low ≥ 1 serving/day).

Statistical Analysis
Analyses were weighted to take into account the oversampling of middle-class households in 1988 and nonresponse to both the 1988 and 2001 surveys. Multiple logistic regression was used to investigate relationships between life-course SEP and obesity status in 1988. Women in the high/high (i.e., most advantaged) life-course SEP category constituted the referent group for all comparisons. Separate tests of main effects for low versus high childhood SEP (after control for adulthood SEP) and for low versus high adulthood SEP (after control for childhood SEP) were also conducted.

Using hierarchical regression models, we added potential confounders and potential intervening variables individually or as a block to isolate their unique contribution to obesity risk. For example, the fully adjusted model describing the association between life-course SEP and obesity risk controlled for potential confounders in the following order: age (model 1), marital status (model 2), alcohol consumption and cigarette smoking (model 3), and childhood food insecurity (model 4). The 3 hypothetical intervening variables (strenuous exercise, low fruit consumption, and low vegetable consumption) were added in model 5. All analyses were performed with SAS, Version 9.12 (SAS Institute Inc, Cary, NC).(n26) Weighted estimates of parameters, variances, and 95% confidence intervals were obtained using either linearization or Jackknife Repeated Replication techniques.(n27) Analyses were restricted to the 679 women with no missing values on study variables.