Bone Mineral Density During and After Lactation: A Comparison of African American and Caucasian Women

Subjects included healthy AA and C mothers ages 21–45 who were postpartum after a singleton pregnancy. Race was self-reported. All participants were breastfeeding, defined as exclusively breastfeeding or using up to one bottle of supplemental formula per day, at the time of the first two study visits.

Exclusion criteria included disorders which could affect bone metabolism, including renal, cardiovascular, pulmonary, malignant, hepatic, hematologic or rheumatologic disease. Other exclusion criteria included fractures or bone surgery within the past 12 months, smoking, history of significant alcohol or drug use, chronic medications including depot medroxyprogesterone acetate (DMPA) (with the exceptions of stable doses of thyroid hormone or vitamin supplements), weight over 130 kg, and BMD Z-score of − 3.0 or lower at the hip or spine. Women who achieved pregnancies with in vitro fertilization (IVF) or other hormonal manipulation were excluded, as were those who had significant complications with the most recent pregnancy or were unable to exclusively breastfeed beginning at birth. Women who became pregnant during the study did not participate in study visits after pregnancy was discovered. The study protocol was approved by the University of Pittsburgh Institutional Review Board and informed consent was obtained from each subject prior to the initiation of study procedures. The study was performed in accordance with the ethical standards and principals of the Declaration of Helsinki.

This was a prospective paired cohort study comparing bone mineral density (BMD) and markers of turnover during lactation and weaning in self-identified AA and C mothers. The primary endpoint was BMD at the lumbar spine (LS), total hip (TH) and femoral neck (FN). Based on reported BMD changes and our prior bone marker results in C women [11], we planned to recruit 25 subjects per group with the goal of at least 20 women per group completing the study.

The study involved four outpatient visits to the Endocrine & Metabolism Research Center at the University of Pittsburgh. The first visit occurred two weeks after delivery (baseline), followed by visits at 12 weeks postpartum (3 months), at 24 weeks postpartum or at the onset of weaning if prior to 24 weeks (6 months), and then 6 months after completely weaning (post-wean). At each visit, a relevant medical history, calcium and physical activity questionnaires, blood and urine samples, breast milk samples, and a dual energy x-ray absorptiometry (DXA) scan of the LS, TH, FN and distal 1/3 of the radius (DR) of the non-dominant limb were obtained. Blood samples were collected for calcium, ionized calcium, albumin, phosphate, PTH (1–84), 25-hydroxyvitamin D (25-OHD), estradiol, and markers of bone turnover [procollagen type 1 N-terminal (P1NP) and C-terminal telopeptide (CTX)] at each visit. Urine calcium, creatinine, and phosphate were measured at every visit on a second morning void. TSH was measured at the 3 month visit. Only subjects who completed both the baseline and 3 month visits and who were still breast feeding at three months were included in the final analysis.

Serum total calcium, ionized calcium, albumin, phosphate, and TSH were measured using standard automated chemistries in the University of Pittsburgh Medical Center (UPMC) Clinical Chemistry Laboratory on the day of collection. Markers of bone turnover, 25-OH D, estradiol and intact PTH were assayed in single batches at the conclusion of the study on blood samples that had been processed and stored at − 80 °C. 25-OH D was measured in the UPMC Clinical Chemistry Laboratory by Liquid Chromatography-Tandem Mass Spectroscopy using Waters Oasis HLB SPE columns (Waters Corporation, United Kingdom, CV = 10%). Intact PTH (1–84) was measured by immunochemiluminometric assay (Quest Laboratories, San Juan Capistrano, CA). P1NP was measured using the IDS-iSYS Intact PINP amino-terminal propeptide of type I procollagen assay (Immunodiagnostics Systems, United Kingdom, CV = 2.2%) and CTX was measured using the Serum Crosslaps ELISA assay (Immunodiagnostics Systems, United Kingdom, CV < 11%). A urine pregnancy test was obtained at each visit prior to the DXA scan (ICON urine hCG).

Spot urine calcium, creatinine and phosphate were measured on the day of each visit using standard automated chemistries in the UPMC Clinical Chemistry Laboratory. Fasting samples were collected when possible, but many patients found it difficult to fast while breastfeeding, so some samples were non-fasting. Fractional excretion of calcium was calculated using FECa = (urine calcium/ionized serum calcium)/(urine creatinine/serum creatinine). Tubular maximum reabsorption of phosphate (TMP) was calculated as described by Payne [29].

DXA scans were all performed by trained study personnel using the same Lunar iDXA machine (GE Healthcare) (in house CV = 1.48%). A single investigator reviewed all scans for comparison to baseline. Calibration of the absorptiometers was done before each study visit using a Lunar phantom. Breasts were shielded for scans done at visits when subjects were lactating.

Baseline characteristics of participants were reported by race as mean (SD) or median (25th percentile, 75th percentile) depending on normality. Comparisons of characteristics between races were conducted using equal or unequal variance t-tests, or a Wilcoxon Rank Sum test as indicated. With close to 50% of the sample having their first child, the distribution of parity in each race exhibited a distinct peak at parity = 1, then tapered off at higher parity values. Subtracting one from parity values, the distributions of parity-1 between races was compared using a negative binomial distribution. This distribution fit better than a Poisson distribution based on AIC.

The primary focus was on changes from baseline when comparing trends over time in BMD. For absolute changes from baseline to 3 months postpartum, 6 months postpartum and 6 months post wean, 4-timepoint longitudinal repeated measures mixed models were used with baseline as the referent. To reduce the number of covariance parameters estimated so that means are better estimated, we conducted goodness-of-fit test using the AIC criterion to identify best-fitting parsimonious covariance structures for each measure and each race. Covariance structures compared were: unstructured, auto-regressive with equal variances at each timepoint, heterogeneous auto-regressive, equi-correlated with equal-variances and heterogenous equi-correlated. Percent changes from baseline were also examined. Percent changes at 3 months, 6 months and 6 months post wean were modeled using 3-timepoint longitudinal repeated measures mixed models with identification of best fitting parsimonious covariance structures using a similar approach. The final repeated measures models included race by timepoint interactions which provided statistical tests for differences in change from baseline at each timepoint between races. For BMD measures over time, changes vs. baseline plus 95% CIs were tabulated at the three post-baseline timepoints. Similar measures were used for analysis of the biomarkers as a secondary endpoint. However, the time-point specific values plus 95% CIs, rather than the changes, were tabulated at each of the post-baseline timepoints. In both cases, the p-values evaluating differences by race at each timepoint are for changes vs baseline based on the relevant repeated measures model. All analyses were conducted using SAS version 9.4 (SAS Institute Inc., Cary). Box plots were created using R version 4.1.2 (2021-11-01).

Forty-five AA and 32 C subjects were enrolled. Of the 29 subjects in each group who completed visits 1 and 2 and were included in the analysis, 16 AA subjects and 20 C subjects completed all four study visits (Fig. 1).

Review of EMR data during the 14-month recruitment period (10/2012–12/2013) indicates that there were 12,230 deliveries at UPMC Magee, a women’s health and maternity hospital from which most subjects were recruited. Of these, 2402 mothers self-identified as AA and 230 did not specify their race. Recruitment of AA participants was considerably more challenging than recruitment of C participants. Although a majority of mothers were breastfeeding at the time of discharge, a smaller percentage for AA new mothers (52%, n = 1252) as compared to non-AA new mothers (74%, n = 7093) were breastfeeding. Furthermore, 638 (27%) AA new mothers received DMPA for birth control prior to discharge, and were therefore not eligible for this study, as compared to 494 (5%) of non-AA new mothers.

AA and C lactating mothers were similar at baseline (2 weeks postpartum) with regard to parity, number of breastfeeding sessions per day, and amount of physical activity per day, as shown in Table 1. AA subjects had a significantly higher BMI than C subjects throughout the study. p values for change in weight over time were nonsignificant for both groups; however, AA subjects’ weight trended up and C subjects’ weight trended down over the study. Baseline average total daily calcium intake through food and supplements was significantly lower in AA mothers than C mothers (955 mg/day vs 1282 mg/day, p = 0.03). Total daily calcium intake decreased in both groups by the final visit, with AA subjects’ calcium intake remaining lower than that of C subjects for the entire study duration.

For this study, breastfeeding was defined as exclusively breastfeeding or using up to one bottle of supplemental formula per day. Most mothers were not using any supplemental formula at baseline. At 3 months, 10 mothers in the AA group and 2 in the C group were using up to one bottle of supplemental formula per day. The study was designed to include women who were still breastfeeding at 6 months and subjects were instructed to contact the study team if they began to wean before 6 months so they could be brought in for the 3rd study visit before they began significant weaning. Despite this design, at the 6 month visit, 7 AA and 2 C mothers were using more than one supplemental bottle per day and one AA mother had fully weaned. By defininion, all subject had fully weaned 6 months before the last study visit. One subject in the AA group and two subjects in the C group could not complete the post wean visit as they were still breastfeeding, each for over 2 years, when the study was closed. Median duration of breastfeeding did not differ between groups [median (25^th percentile, 75th percentile): AA = 9.0 months (6.5, 11.0); C = 8.0 months (6.0, 12.5); p = 0.80].

All subjects had resumed menses at the time of weaning. On average menses resumed at 5 months postpartum for both groups (AA range 2–14 months, C range 1.5–18 months).

At baseline, total 25-OH Vitamin D levels were lower in AA subjects than in C subjects (28.72 vs 44.03 ng/ml, p < 0.0001). PTH was marginally, but not significantly higher in AA subjects compared to C subjects at baseline. Serum phosphate was lower in AA subjects than in C subjects which corresponded with a trend for lower ratio of tubular maximum reabsorption of phosphate to glomerular filtration rate (TMP/GFR) in the AA group. There were no statistically significant differences between groups at baseline with regard to total calcium, ionized calcium, creatinine, or albumin. However, FECa was lower in AA participants as compared to C participants (Table 1).

Estradiol levels were significantly higher in AA women as compared to C women and this difference remained significant when adjusted for BMI. There were no statistically significant differences between groups in baseline markers of bone turnover as measured by CTX and P1NP (Table 1).

At baseline AA participants had significantly higher absolute BMD at all sites compared to C participants (Table 1).

AA Group: BMD decreased from baseline at the LS, TH, and FN in the AA group at 3 and 6 months and then returned to baseline at all three sites after weaning (Table 2, Fig. 2A–C). Distal radius (DR) BMD remained near baseline at all time points for the AA group (Table 2, Fig. 2D).

C Group: BMD also decreased from baseline at the LS, TH, and FN in the C group at 3 and 6 months and then returned to baseline at the LS and TH after weaning (Table 2, Fig. 2A, B). There was a persistent loss of BMD at the FN at 6 months post-wean in the C group (Fig. 2C). DR BMD decreased by 1.08% (p = 0.01) in the C group at 3 months (Table 2, Fig. 2D) and did not differ from baseline at 6 months and post-wean.

Racial Comparison: The decrease in BMD at the LS did not differ between the groups. However, at 6 months postpartum, the decrease in BMD at the TH was greater for the C group compared to the AA group (Table 2 and Fig. 2B). The FN was the site of greatest absolute bone loss in the C group with a 5.24% loss at 6 months postpartum (p < 0.0001) and a persistent loss of 2.64% at 6 months post-wean (p = 0.0002) (Fig. 2C). In contrast, FN BMD in the AA group was only 1.75% below baseline at 6 months postpartum (p = 0.007) and had returned to baseline by 6 months post wean (p = 0.66). The p-values were significant for racial differences at the FN at 3 months and 6 months (Table 2, Fig. 2C) and for change over time (p = 0.02).

BMI was positively associated with FN BMD (p = 0.0005), TH BMD (p = 0.001) and DR BMD (p = 0.001) but not significantly associated with LS BMD. Significance of racial differences in BMD did not change after adjustment for BMI.

Estradiol levels rose over time in both groups (Table 3). Despite higher baseline estradiol levels, there was also a significantly greater increase from baseline estradiol levels in AA women compared to C women 6 months post wean (178.53 vs 75.58 pg/ml, p = 0.03). Adjustment for BMI did not change significance of racial differences in estradiol levels.

Vitamin D levels decreased from baseline in both groups, though the difference between groups only met significance at the 3 month timepoint (p = 0.04). BMI was negatively associated with Total 25-OH Vitamin D (p = 0.003). Adjustment for BMI did not change significance of racial differences in vitamin D. Overall PTH was stable throughout all timepoints the study with no racial differences.

Bone resorption, as measured by CTX was stable during lactation and then decreased to 30% below baseline by 6 months after weaning in both groups (change from baseline p = 0.005 for AA and p < 0.0001 for C, Fig. 3A). There was no overall difference between the groups over time.

Bone formation, as measured by P1NP, increased significantly during lactation in both groups. After weaning, P1NP levels returned to baseline in the AA group but fell to 32% below baseline in the C group (p < 0.0001 for C, p = 0.03 for racial difference, Fig. 3B).