To test the hypotheses above, 25 postmenopausal grandmothers and 20 control subjects were recruited from greater Kingston, Jamaica. While Jamaica has a population of around 2 million, Kingston is the largest urban area on the island, harboring a socioeconomically diverse population of around 800,000 inhabitants. Approximately 93% of the island’s population claims African Caribbean descent, with the remainder a mixture of European, Native Caribbean (e.g., Taino), South Indian, Chinese, and Southwest Asian ancestry.

We limited our study to postmenopausal Jamaican grandmothers for several reasons. The primary reason is that postmenopausal grandmothers are a culturally salient group of caregivers in Jamaica. This makes the study of them culturally normative rather than rare or unusual. Another reason is that a female’s reproductive status (pre- vs. postmenopausal) affects oxytocin and prolactin levels²⁵, meaning that by focusing on strictly postmenopausal women we can remove the potentially confounding effects of female reproductive status on these hormones. To further restrict variation among subjects in this study, we limit our study to postmenopausal grandmothers aged 50-65 who do not have a biological child less than 10 years of age and who co-reside (live in the same dwelling) with at least one putative biological grandchild aged 5 or younger. We sought control subjects of comparable age and socioeconomic backgrounds, but defined controls by virtue of not living with a putative biological grandchild aged 5 or younger. We recruited groups of comparable age and socioeconomic backgrounds by checking for similar age and educational backgrounds of participants during the recruitment phase. Menopause was operationalized as a self-reported cessation of menstruation during the previous 12 months.

According to 2004 data (www.worldbank.org), average lifespan in Jamaica is 70.8 years, meaning that many women reach these upper ages. However, the population has experienced an increase in the incidence of obesity and metabolic diseases such as Type 2 diabetes and hypertension. Accordingly, we included in our sample about half of subjects who have been diagnosed with Type 2 diabetes and/or hypertension. Finally, although completed fertility in Jamaica has fallen to 2.4 children, we encountered few older women who had not previously had children and observed that Jamaican grandmothers continue to serve important roles as caregivers of their grandchildren.

We recruited grandmothers and control women by word of mouth and at two churches in greater Kingston. One of these churches was an Anglican church in urban Kingston, and the other a Seventh Day Adventist church on the rugged, mountainous outskirts of town. These churches were chosen because one of us (MS-V) maintains active and longstanding ties with one, and because of ties to research personnel in the other case. Subjects were remunerated either $3500 Jamaican (approximately $50 U.S.: grandmothers) or $2000 Jamaican (approximately $30 U.S.: controls). Ethical approval for the study was given by the University of the West Indies Faculty of Medical Science/University Hospital Ethical Committee and University of Nevada, Las Vegas institutional review board. Written informed consent was obtained from participants.

The study design was built around testing both within- and between-subject effects of caregiving on hormones. To test within-subject effects, we performed all study procedures with grandmothers on two days: one day in which the grandmother had spent the previous four hours caring for her youngest grandchild, and a second day in which the grandmother had not been caring for this or other young grandchildren the previous four hours. Potential within-subject effects of caregiving on hormones were expected to be shown on the caregiving day. To test between-subject effects, we recruited control women of similar ages, socioeconomic variation, and health status. All procedures with control participants took place on one day. Data from controls were then compared with the two-day average day of grandmothers in addition to supplemental analyses contrasting controls with grandmothers on caregiving and non-caregiving days respectively.

All study procedures took place either in a subject’s home or at one of the same two churches from which participants were recruited. This feature of the study design was meant to capture more naturalistic effects of grandmothering on hormones. In all cases where grandmothers participated at a church, they lived within short walking distances. Since the church serves as a central social activity for many Jamaican women, including those in the sample, conducting procedures in the church enabled use of a familiar environment.

All study procedures took place between 1200 and 1800 hours. One reason for limiting procedures to these times was to reduce potential circadian effects on hormone levels (e.g., most declines in cortisol levels occur within the first few hours after waking). Another reason was to conduct study procedures after grandmothers had been awake for enough hours to have either spent the previous four hours providing care or not of a grandchild. All subjects participated between 28 May 2008 and 18 July 2008. The order of procedures entailed subjects providing informed consent, biological samples, and then responding to interview items.

A central methodological feature of this study was use of minimally invasive sample collection. Saliva samples were collected in order to measure cortisol levels; finger prick blood spot samples were collected to measure prolactin levels; and urine samples were collected to measure oxytocin and vasopressin levels. Procedures of sample collection have been previously published (see ²²). In brief, these procedures entailed collecting biological samples as follows: approximately 1.0 ml of unstimulated saliva was provided through a straw into a cryovial tube; approximately 4 drops of blood were dropped upon filter paper after wiping the finger with an alcohol wipe and pricking it with a safety lancet; and urine samples were provided in specimen collection cups then transferred into polystyrene tubes for long-term storage. After samples were collected, they were kept in a cooler along with freeze packs before transported to a -20C freezer at the University Hospital of the West Indies for storage until shipped back to the U.S. for assay. In addition to collecting these three biological samples during each study visit, we also measured blood pressure using a portable blood pressure monitor (Omron HEB-773AC), measured height and weight to calculate body mass index (BMI: kg/m²), and engaged in semi-structured interviews.

Interviews were conducted with each control subject and each grandmother on her “non-caregiving” day. Interviews included questions covering basic sociodemographic information such as age, socioeconomic status, and reproductive history. Educational attainment was scored as 1 (none); 2 (primary school); 3 (secondary school); or 4 (post-secondary school). A measure of wealth was indexed as the total number of a possible 11 household items (e.g., refrigerator, cell phone, stereo/VCR/DVD equipment, motor vehicle) present in a subject’s household.

To provide further social context on Jamaican grandmothers’ lives and caregiving, we asked open-ended questions concerning household structure, broader behavioral context of caregiving, and emic views of caregiving. Examples of these open-ended questions include: How do you feel about providing care for your grandchild(ren)? and What types of support do you provide? We report some of the qualitative data stemming from these interviews in the Discussion section. Interviews typically lasted between 30 and 60 minutes. Interviews were either conducted by one of us (PG) or trained Jamaican research assistants.

Details concerning salivary, blood spot, and urinary hormone assays have been previously published in²². These methods included solid phase extraction for sample purification (SepPak C18 cartridges, Waters, cat# WAT023590) prior to assay. Samples were 1 ml aliquots of urine for each assay and after solid phase extraction the samples dried, resuspended in assay buffer and were applied onto the EIA plates in 100 µl (Assay Designs,Ann Arbor, MI) . Results of the prolactin assays yielded inter-assay coefficients of variation of 13.3% and 8.9% for low and high controls, respectively, and an intraassay coefficient of variation of 8.0%. Interassay coefficients of variation for cortisol were 8.2% and 2.3% for low and high controls, respectively, and an intrassay coefficient of variation of 11.8% Oxytocin coefficients of variation were 6.4% for intrassay and 14.1% for interassay. Vasopressin coefficients of variation were 4% for intrassay and 14% for interassay.

K-S tests indicated that cortisol, vasopressin, and oxytocin levels, but not prolactin levels, differed significantly from a normal distribution (p<0.05). Thus, cortisol, vasopressin, and oxytocin concentrations were log(10) transformed. After log transformation, these hormone concentrations no longer deviated from a normal distribution (p>0.05). Two prolactin values that were > 3SD above the mean were removed. No log transformed cortisol, oxytocin, or vasopressin levels were > 3SD above the mean.

Hormone analyses included five potential confounding variables: age, BMI, current use of daily medications, time of day, and the location (home vs. church) at which study procedures took place. Initial bivariate analyses relied on either t tests or ANOVA to determine whether these variables were associated with hormone concentrations (Table 1). Results of these bivariate explorations of potential confounding variables suggested several variables that should be controlled. Those variables exhibiting relationships at p<0.10 were retained in multivariate models designed to test for effects of caregiving on hormone concentrations. ANCOVA was used for all multivariate models. To test within-subject effects of caregiving on hormones, repeat measure ANOVA was used with hormone concentrations used as the repeat dependent measures and any covariates included. To test for between-subject effects, we used grandmothers/controls as the independent variable, included relevant covariates, and a subject’s hormone concentrations were used as the dependent variables. The average of a grandmother’s two hormone concentrations (e.g., cortisol on the two days measured) was used for grandmothers in these between-group analyses. Posthoc analyses (not shown) revealed that results did not noticeably differ if controls were compared with grandmothers on caregiving and non-caregiving days, respectively, rather than the average of grandmothers’ hormone levels.

Figure 1

Table 1. Univariate analyses of hormones and potential confounding variables.

We recruited 45 women aged 50-67 to participate. The group of 25 grandmothers did not differ significantly from the 20 “control” women in age, BMI, education, household material items, or employment status (see Table 2). The completed fertility was not significantly different between the two groups of women either. Each woman in the sample had born at least one child. Commonly, women had children fathered by different men. In one of the more extreme examples, a grandmother reported having 8 children fathered by 5 men; more typical was a woman having 3 children by 2 men. The only significant difference between the groups of women was that grandmothers were less likely to be involved in a current relationship with a man than were control women (t=3.259, df=34.25, p=0.003). Formal marriage was relatively rare, with partnerships more equivalent to live-in common-law unions.

Figure 2

Table 2. Sociodemographic data (Mean ± SD) of participants.

Most (N=16) grandmothers served supplemental roles to a parent or parents rather than as primary caregivers (N=9). In a few cases, grandmothers lived with their son and grandchildren, but in the vast majority of cases grandmothers lived with a daughter, sometimes the daughter’s partner, and the daughter’s children (hence emphasizing maternal grandmothering). Grandmothers reported frequent patterns of caregiving that emphasized their roles spending time with grandchildren (especially while the children’s parents worked); helping defray costs of material items such as school, food and medications; helping protect their grandchildren from danger (e.g., keeping them off rough urban roads); and acting as indulgent caregivers compared to the children’s parents (e.g., holding a crying grandchild).

There were no differences between grandmothers and controls in the proportion of women reporting a diagnosis of Type 2 diabetes and/or hypertension in the study. The average blood pressures (Table 3) were quite high, but also consistent with broader Caribbean data. Interestingly, no women indicated taking hormone therapy.

Figure 3

Table 3. Blood pressure and hormone data (Mean ± SD) of participants.

Hormone concentrations are reported in Table 3. In multivariate models testing for within-subject effects of caregiving on cortisol, the overall model (F=3.701, df=23, p=0.032) was significant, but caregiving day (F=.594, p=0.449) was not related to cortisol levels, though use of medications was (F=5.472, p=0.028). In testing between-subject effects (controls vs. grandmothers), the overall model was not significant (F=.704, df=42, p=0.500). Use of daily medications (F=.877, p=0.565) and caregiving (F=.337, p=0.565) were not associated with cortisol levels.

In multivariate models testing for within-subjects effects of caregiving on prolactin, the overall model was not significant (F=.955, df=21, p=0.423), and BMI (F=1.347, p=0.259), age (F=.175, p=0.680), and caregiving day (F=1.143, p=0.297) were also not significant predictors of prolactin. In testing between-subject effects, the overall model (F=3.235, df=40, p=0.034) was significant. BMI (F=5.760, p=0.022) was significantly associated with prolactin, age exhibited a tendency (F=2.969, p=0.094) toward an association with prolactin, and caregiving (grandmothers vs. controls) was not associated (F=.368, p=0.548) with prolactin levels.

In multivariate models testing for within-subject effects of caregiving on oxytocin, the overall model was not significant (F=1.785, df=16, p=0.166), nor were age (F=.997, p=0.333), location (F=.604, p=0.449) or caregiving day (F=.125, p=0.728). In testing between-subject effects (grandmothers vs. controls), the overall model neared significance (F=2.853, df=38, p=0.053). Location (F=.632, p=0.432) and caregiving (F=.055, p=0.816) were not significantly, though age (F=5.944, p=0.021) was significantly associated with oxytocin levels.

In multivariate models testing for within-subjects effects of caregiving on vasopressin, the overall model was not significant (F=.479, df=21, p=0.622), nor were location (F=..365, p=0.552) or caregiving day (F=.010, p=0.920). In testing between-subject effects, the overall model was significant (F=5.653, df=41, p=0.007). Location was not significantly associated with vasopressin (F=2.166, p=0.149), but caregiving was (F=6.080, p=0.018).

.

Figure 4

Figure 1. Vasopressin levels according to grandmothering status

We thank all of the women who participated in this study. We thank the field research team of Alex Jackson, Anniki Witter, Nicholas Vaughan, and Ann-Marie Graham. We are grateful for labwork conducted by Fred Hadi and Dan Wittwer. For comments on earlier drafts of the manuscript, we thank Michelle Escasa, Isal Guillermo, Chandler Marrs, Fred Kuch, Heidi Manlove, Robin Nelson, Tom Steiner, Sari van Anders, Sharon Young, and Toni Ziegler. For financial support, we are indebted to UNLV (Research Development Award to PG) and in part assay support from the Wisconsin National Primate Research Center, RR000167, a component of the National Institutes of Health.