The epidemiologic evidence linking stress and poor pregnancy outcomes has been intriguing, although somewhat inconsistent. While several studies have shown a link between maternal self-reported stress and low birth weight or preterm birth,^{15, 17–23} others have failed to demonstrate an association.^24–28 Similarly, it also has been hypothesized that maternal stress may play a role in explaining the racial disparities seen in preterm birth rates in the United States.²⁹ Yet, the available data have not consistently established disparities in stress as the underlying etiology for disparities in preterm birth.

Several factors may explain the inconsistent associations reported between maternal stress and preterm birth. Much of the research conducted to date relies heavily upon self-report measures of stress. In addition, numerous measures of maternal stress have been used with little consistency in the operational definitions across studies. Perhaps most importantly, existing research largely has focused upon acute stress with little emphasis upon chronic stress.³⁰ This is important because chronic stress is a biologically plausible mechanisms linking stress with preterm birth.

The sample for the current analysis includes the IFS participants from Cook County in 2005. At the time of their yearly interview in 2005, IFS study participants answered several additional survey questions regarding self-reported stress and provided a blood specimen. Given the community setting of the IFS study, blood specimen collection was conducted in participants’ homes.³⁸ Specifically, IFS interviewers were trained to collect a dried blood spot specimen from a finger stick. The blood spots were collected on standardized filter paper commonly used for neonatal screening (No. 903, Schleicher and Schull, Keene, NH), allowed to dry for four hours, and then returned to and stored in a central l aboratory at −30 C until batched analysis.

Blood spots were analyzed for C-reactive protein (CRP) and Epstein Barr virus (EBV) antibodies using standardized enzyme immunoassay protocols developed for use with dried blood spot samples.^39–40 After discs of whole blood were punched from the dried filter paper cards and eluted overnight, the eluate was added to microtiter wells. The quantity of CRP in each sample was determined based on comparison with calibrator values (Dako Corporation, product no.X0923) using a spline-fit standard curve. The EBV antibody protocol utilized an adaptation of a commercially available kit for measuring EBV viral capsid antigen (VCA) IgG antibodies in serum (KC Junior, BioTek, Winooski, VT). The assay quantifies IgG antibodies against the p18 polypeptide, a marker protein containing the immunodominant epitopes of the viral capsid antigen complex. Control and sample values are interpolated from a standard curve using a linear data reduction protocol and reported in ELISA units (Soft-Max Pro, Molecular Devices, Sunnyvale, CA). Previous validation studies indicate a high degree of correlation between results obtained from paired blood spot and serum samples for both CRP³⁹ and EBV antibodies. In the present analysis, subjects with EBV values less than 20 ELISA units were excluded from EBV-related analyses, given that this result is consistent with a non-EBV exposed individual.

Psychosocial factors were grouped into four categories, based on a previously reported conceptual model of chronic stress.^{11, 19} This model posits four domains that are particularly relevant to understanding the actual internal stress experienced by an individual: (1) external stressors (life events, hardships in the home environment, hardships obtaining food, hardships obtaining medical care, having a child with chronic illness in the home, and home crowdedness); (2) enhancers of stress (depression, mental health, and drug or alcohol use); (3) buffers against stress (social support, community group involvement, and coping skills); and (4) perceptions of stress (perceived economic hardship, self-rated health, perceived neighborhood safety, and discrimination). The questions and validated scales from the IFS questionnaire are shown in Table 1.

The relationships between demographic and psychosocial variables and levels of CRP and EBV were investigated. Patients were stratified into “low” or “high” stress categories, as previously described, based on their responses to the stress scales.¹⁷ The serum levels of biomarkers were evaluated for normal distribution using the Kolmogorov-Smirnov test. Continuous variables that were normally distributed were compared using the student t-test and those that were not normally distributed were compared using the Mann-Whitney U test. Pearson analysis was used to assess correlation. All tests were two-tailed and P < .05 was used to determine significance. Data analysis was performed using SPSS for Windows, version 16.0 (SPSS Inc, Chicago, IL).

The CRP levels, which were not normally distributed (P < 0.01), ranged from .02 to 35.54 mg/L with a median (interquartile range) of 1.8 (0.6–5.5) mg/L. Four EBV samples were less than 20 ELISA units and therefore excluded from the analysis. The remaining specimens (N = 183) yielded EBV levels that were normally distributed (P = 0.12) and ranged from 22.69 to 303.39 ELISA units, with a mean of 136.84 ± 71.70 ELISA units. EBV and CRP levels were found to be significantly correlated (p<0.001).

Elevated CRP levels have been associated with self-reported stress.^36–37 Similarly, elevated CRP levels have been associated with preterm birth.^{31, 34–35} Yet, there have been no prospective studies that have linked stress, CRP levels, and preterm birth, and the relevance of CRP as a biomarker of stress relevant to reproductive-age women and to adverse obstetric outcomes has not been established. Indeed, in this prospective study, CRP levels were not strongly associated with multiple domains of an individual’s experience of stress.

However, EBV antibody levels were significantly associated with measures of stress. This relationship is of particular interest, given that the relationship of EBV antibodies with chronic stress has not been well studied in reproductive-age or pregnant women. EBV antibodies, however, have been widely studied among other populations and shown to be associated with negative life events, academic stress, strained social relationships, and emotional distress.^{8, 13, 32–33, 41–44} Indeed, in one meta-analysis, EBV antibodies were identified as one of the most consistent and strongest measures of stress.⁹

In this study of low-income, predominantly African-American women, we found elevated EBV antibody levels to be most strongly associated with elevated perceived stress and experiences of discrimination. Measures of perceived stress and discrimination have been associated with increased rates of low birth weight infants and preterm birth in African-American women. For example, Collins et al. reported that African-American mothers’ perceptions of increased stress were significantly associated with increased odds of having a very-low birth weight (VLBW) child. Additionally, it was noted that African-American women who reported higher exposure to racism had significantly increased odds for very low birth weight (VLBW) infants even when controlling for demographic, biomedical, and behavioral variables.⁴⁵ The association of racial discrimination with adverse pregnancy outcomes, as well as with higher levels of EBV antibodies, is of particular relevance given the known racial disparity in preterm birth, the lack of an accepted etiology of the disparity, and the inability of traditional socioeconomic variables (e.g. economic differences) to fully account for this disparity.

This study also provides evidence that blood specimens, obtained in a community setting with a high degree of subject acceptance, can allow analysis of potentially relevant stress biomarkers. It has previously been shown that community-based research may more effectively overcome issues of trust, ease of access, and follow-up than hospital-based specimen collection, particularly in low-income populations.^46–48 Community-based data collection may achieve higher rates of participation and retention and therefore diminish selection bias that may be associated with hospital-based data collection.

Limitations of our study should be noted. The size of the study population may have limited the ability to elucidate associations between biomarkers and survey measures of stress. Also, the detection of an association depends upon the ability to discern the stress that an individual is experiencing, and the best method or instrument to assess chronic stress remains uncertain. It may be that stress is best defined not by measurement from single instruments, but by a more elaborate combination of measures f rom different domains and instruments. Not all measures of stress used in surveys may be equally valid or reliable for different populations. For example, stressors experienced by low-income women living in an urban setting may need to be assessed differently than for women who have other socioeconomic or racial backgrounds. Nevertheless, the measures used in this study are well established and assess different domains of stress. Finally, women in the study all had a significant degree of economic deprivation which reduced any potential confounding by economic status. This population was chosen because chronic stress was likely to be more prevalent and greater in magnitude. Yet, it is also possible that the choice of this population limited the possibility of finding an association, as the overall levels of the stress biomarkers were higher and the distribution of these biomarkers was narrower than in a random sample of the entire population. In this case, a “ceiling effect” could have impaired the ability to discern associations between chronic stress and the biomarkers under study.