Abstract
Background
Analysis of cytokines and growth factors in human milk offers a noninvasive approach for studying the microenvironment of the postpartum breast, which may better reflect tissue levels than testing blood samples. Given that Black women have a higher incidence of early onset breast cancers than White women, we hypothesized that milk of the former contains higher levels of pro-inflammatory cytokines, adipokines, and growth factors.
Methods
Participants included 130 Black and 162 White women without a history of a breast biopsy who completed a health assessment questionnaire and donated milk for research. Concentrations of 15 analytes in milk were examined using two multiplex and 4 single-analyte electrochemiluminescent sandwich assays to measure pro-inflammatory cytokines, angiogenesis factors, and adipokines. Mixed-effects ordinal logistic regression was used to identify determinants of analyte levels and to compare results by race, with adjustment for confounders. Factor analysis was used to examine covariation among analytes.
Results
Thirteen of 15 analytes were detected in ≥25% of the human milk specimens. In multivariable models, elevated BMI was significantly associated with increased concentrations of 5 cytokines: IL-1β, bFGF, FASL, EGF, and leptin (all p-trend < 0.05). Black women had significantly higher levels of leptin and IL-1β, controlling for BMI. Factor analysis of analyte levels identified two factors related to inflammation and growth factor pathways
Conclusion
This exploratory study demonstrated the feasibility of measuring pro-inflammatory cytokines, adipokines and angiogenesis factors in human milk, and revealed higher levels of some pro-inflammatory factors, as well as increased leptin levels, among Black as compared with White women.
Keywords: Human milk, breast cancer risk, prevention, race
Introduction
Increased lifetime duration of lactation in developed countries has been associated with a modest lowering of breast cancer risk overall, with a potentially stronger protective effect on risk for aggressive, early onset estrogen receptor (ER)-negative or triple-negative breast cancers (1–4). Differences in exposures that might increase risk of ER-negative tumors, such as increased parity, less breastfeeding and greater body mass index (BMI) may contribute to the higher age-adjusted incidence of ER-negative breast cancers among Black compared with White women (5). Findings from the AMBER consortium show that an early age at first birth, particularly without breastfeeding, may contribute to an elevated risk of triple negative breast cancer among Black women (3, 4). Further, a meta-analysis found that premenopausal obesity is associated with increased risk of triple negative breast cancer (8). Understanding the interplay of parity, lactation, BMI, and early onset breast cancer is potentially important for risk assessment and prevention, but developing approaches to pursue this research is challenging.
We hypothesized that factors associated with early onset breast cancer, such as obesity and race, might be related to increased levels of cytokines, angiogenesis factors, and hormones in the post-partum period and that such factors are measurable in milk. Given that previous studies have shown that levels of specific cytokines in human milk and serum differ (9–11), measurement of these components in milk may provide a better reflection of the breast microenvironment than analysis of blood. Further, limited data suggest that inflammatory protein levels differ between affected and unaffected breasts of women with unilateral breast cancer (12), supporting the idea that milk may provide information about cancer risk that pertains specifically to the sampled breast. Several studies have measured cytokines in milk (13–14), including sample sizes ranging from 5 to 182 specimens, but none has focused on comparing analyte concentrations in Black and White women. In the current exploratory study, we demonstrate the feasibility of comparing cytokine and growth factor levels in human milk by participant characteristics to develop hypotheses that may account for racial differences in risk of early onset breast cancers.
Methods
Study population
Participant recruitment, milk collection and processing are described in detail elsewhere (13–15). Briefly, from 2007-2013 lactating women aged 18 years or older were recruited to donate human milk samples, which were frozen and archived at the University of Massachusetts-Amherst Breastmilk Laboratory. Women were recruited for milk donation through websites, lactation consultants, or, for those living within 20 miles of the laboratory, local advertisement or fliers. For the purposes of this analysis, we selected a total of 292 samples, 130 donated by Black women and 162 by White women, all without a history of a breast biopsy or breast cancer. Women provided informed consent and a self-completed questionnaire related to medical history, lactation and breast cancer risk. The study was approved by the Institutional Review Boards of the University of Massachusetts Amherst and the National Cancer Institute.
Enrolled women were provided with pre-labeled glass bottles or BPA-free plastic milk bags and asked to express the full breast contents of each breast upon waking into separate containers by either hand expressing or using their own breast milk pump. For participants living with 30 miles of the University of Massachusetts-Amherst, milk was collected at their home and transported to the laboratory in a cooler. All other participants were provided with coolers, ice packs and prepaid labels for home pick-up and express (overnight) shipment of their milk to the laboratory.
Laboratory methods
Milk samples were processed immediately upon arriving at the laboratory. Cells were separated from the milk after centrifugation resulting in cell-depleted milk diluted 1.5x with phosphate buffered saline (PBS). Diluted milk samples were stored at −20°C.
Cytokines and growth factors were measured using two multiplex and four single-analyte assays from Mesoscale Discovery (MSD, Gaithersburg, MD). A custom-ordered Human V-PLEX Proinflammatory Panel 1 included assays for five cytokines: interleukin-6 (IL-6, lower limit of detection (LLOD): 0.14 pg/mL), interleukin-8 (IL-8, LLOD: 0.12 pg/ml), tumor necrosis factor-alpha (TNF-α, LLOD: 0.09 pg/mL), interleukin-1 beta (IL-1β, LLOD; 0.07 pg/mL), and interferon-gamma (IFN-γ, LLOD: 0.74 pg/mL). A custom-ordered Human V-PLEX Angiogenesis Panel 1 included assays for six growth factors: vascular endothelial growth factor C (VEGFC, LLOD: 13.81 pg/mL), vascular endothelial growth factor D (VEGFD, LLOD: 3.26 pg/mL), tyrosine kinase with immunoglobulin-like and EGF-like domains 2 (TIE-2, LLOD: 21.69 pg/mL), fms-related tyrosine kinase 1 (FLT1, LLOD: 1.53 pg/mL), placental growth factor (PLGF, LLOD: 0.42 pg/mL), and basic fibroblast growth factor (bFGF, LLOD: 0.15 pg/mL). Single analyte assays were used for leptin (LLOD: 93.96 pg/mL), adiponectin (LLOD: 0.008 ng/mL), Fas ligand (FASL, LLOD: 0.415 pg/mL), and epidermal growth factor (EGF, LLOD: 0.075 pg/mL). Assays were performed according to the manufacturer’s instructions. Each 96-well plate included 38 human milk samples, eight standards in duplicate, and 4 control wells. Eight-point standard curves generated with a 4-parameter logistic regression model were used to provide calculated concentrations for each analyte. The standard curves demonstrated adequate consistency across plates (data not shown). One sample was repeated in duplicate on all 8 plates, and one of two other samples was included in duplicate on each of four plates. Plates (n = 8) were read on the SECTOR® Imager 2400A (MSD). MSD uses electrochemiluminescence technology with a tagged double-antibody sandwich assay to detect each analyte of interest.
Statistical Analysis
All 15 analytes were measured in technical replicates and analyzed as both continuous and categorical variables based on the proportion of samples within the detectable range. Assay precision was assessed using generalized linear models with correction for analyte levels within plates to calculate coefficients of variation (CVs) and intraclass correlation coefficients (ICCs). Final calculated concentrations of analytes were further categorized based on the proportion of individuals with measurements less than the LLOD as follows: analytes for which > 75% of samples yielded measurement above the LLOD were categorized into quartiles (adiponectin, IL-8, EGF, FLT1, PLGF, VEGFD, FASL); analytes for which 50-75% of samples measured above the LLOD were analyzed in four categories, defined as less than LLOD and tertiles of detectable measurements (bFGF, IL-6, IL-1-beta, Leptin, TNFα); and analytes for which 25-50% of sample values were above the LLOD were analyzed as three categories: less than LLOD and below and above the median (IFNy). Most (121 of 130; 93%) milk samples donated by Black women were mailed to the laboratory, whereas 129 of 162 (80%) White women residing within a one-hour drive from the laboratory had their samples collected at home and brought to the laboratory. The primary data analysis proceeded with a pooled sample of 292 women, based on sensitivity analyses that demonstrated that associations were independent of method of donation, irrespective of race (data not shown).
Ordinal logistic regression models with a random effect for plate were used to test for associations between maternal demographic characteristics (predictor variables) and analytes in human milk, which were categorized after taking into account detectability as described in the previous paragraph. Consistent with an exploratory analysis, we considered p<0.05 statistically significant. We present unadjusted and adjusted model results. Adjustment variables were selected by forward stepwise selection, including race plus all variables with p-values less than 0.4 retained in final models. We used factor analysis, with varimax rotation, to explore the relationships of cytokines, angiogenesis factors, and adipokines in human milk. We computed analyte factor scores for each participant, categorized them into quartiles, and evaluated their relationship with maternal and breastfeeding characteristics using ordinal logistic regression, controlling for plate. Adjustment variables were selected by forward stepwise selection as described above.
Descriptive statistics and ordinal logistic regression analyses were performed using STATA 13 (College Station, TX). CVs, ICCs and factor analyses were computed using SAS version 9.3 (Cary, NC).
Results
Population
Demographic and health characteristics of the 292 women are summarized in Table 1, stratified by self-reported race for 130 Black, and 162 White women. Black women tended to have a higher BMI, and greater parity. White women more frequently reported a history of a first-degree relative with breast cancer and ever smoking. Black and White women did not differ in age distribution, the age of their nursing baby, whether their menses had returned since delivery, their age at first birth, or their total lifetime days of lactation. Black women were more likely than White women to have shipped their milk samples (vs. local collection - see Table 1). Demographic characteristics stratified by collection method are shown in Table S1. Women who shipped samples had higher median lactation duration (59% vs. 43% over 180 days), and lower rates of ever smoking (13% vs. 42%).
Table 1.
Demographic and health characteristics of human milk donors
| Black n=130 | White n=162 | ||||
|---|---|---|---|---|---|
| n | % | n | % | p-value* | |
| Age, y | |||||
| <25 | 17 | 13% | 24 | 15% | 0.4 |
| 25 - <30 | 34 | 26% | 32 | 20% | |
| 30 - <35 | 47 | 36% | 53 | 33% | |
| 35 - <40 | 26 | 20% | 40 | 25% | |
| 40+ | 6 | 5% | 13 | 8% | |
| Body Mass Index (BMI), kg/m2 | |||||
| <25 | 59 | 45% | 89 | 55% | 0.05 |
| 25 - <30 | 35 | 27% | 47 | 29% | |
| ≥30 | 36 | 28% | 26 | 16% | |
| Baby’s age, days | |||||
| 21 - <187 | 56 | 43% | 86 | 53% | 0.09 |
| 187 - 1500 | 74 | 57% | 76 | 47% | |
| Menses have returned since giving birth | |||||
| No | 65 | 50% | 93 | 57% | 0.13 |
| Yes | 64 | 49% | 64 | 40% | |
| Missing | 1 | 1% | 5 | 3% | |
| First degree relative with breast cancer | |||||
| No | 122 | 94% | 139 | 86% | 0.03 |
| Yes | 8 | 6% | 23 | 14% | |
| Age at menarche, y | |||||
| <12 | 43 | 33% | 42 | 26% | 0.38 |
| 12 - <14 | 68 | 52% | 84 | 52% | |
| 14+ | 15 | 12% | 30 | 19% | |
| Unknown/missing | 4 | 3% | 6 | 3% | |
| Number of pregnancies | |||||
| 1 | 39 | 30% | 65 | 40% | 0.02 |
| 2 | 33 | 25% | 46 | 28% | |
| 3 | 28 | 22% | 35 | 22% | |
| 4+ | 30 | 23% | 16 | 10% | |
| Ever smoke cigarettes | |||||
| No | 114 | 87% | 100 | 62% | 0.0001 |
| Yes | 16 | 12% | 62 | 38% | |
| Age at first birth, y | |||||
| 15 - <25 | 46 | 35% | 41 | 25% | 0.17 |
| 25 - 43 | 84 | 65% | 121 | 75% | |
| Type of milk collection | |||||
| Local | 9 | 7% | 129 | 80% | 0.0001 |
| Shipped | 121 | 93% | 33 | 20% | |
| Total lifetime days of breastfeeding | |||||
| Q1 21 - <165 | 26 | 20% | 47 | 29% | 0.30 |
| Q2 165 - <390 | 35 | 27% | 38 | 24% | |
| Q3 390 - <753 | 38 | 29% | 38 | 24% | |
| Q4 753 - 3900 | 31 | 24% | 39 | 23% | |
Chi-square p-value
Assay performance
Coefficients of variation (CVs) and intraclass correlation coefficient (ICCs) for 32 repeated samples run across all plates are shown in Table S2. CVs for 8 of the 15 analytes were below 15%, ranging from 2.97% to 14.31% (FASL, EGF, IL-8, FLT1, PLGF, VEGFD, leptin, and adiponectin).CVs and ICCs were similar in measurement of technical replicates across all 292 women in the study (Table S3). TIE-2 and VEGFC were dropped from further analysis due to low detectability (<25%), high CVs and lower ICCs.
For most analytes, the median and range of concentrations, as well as the percent of samples with detectable concentrations, were similar for both locally collected and shipped milk samples (see Table 2). There was only one statistically significant difference: compared with locally collected milk, shipped milk had a higher median value for leptin (p = 0.03), but similar levels of detectability (70% and 80%, respectively, p = 0.07). Because results from shipped and locally collected samples were largely similar, we included results from all 292 women in this exploratory analysis.
Table 2.
Concentrations* of select analytes in human milk
| Locally collected (n = 138) | Shipped to laboratory (n = 154) | |||||
|---|---|---|---|---|---|---|
| Analytes | Median | IQR | Detectable | Median | IQR Detectable | |
| Proinflammatory panel | ||||||
| Interferon gamma (IFN | 0 | 0 - 6.22 | 37% | 0 | 0 - 7.00 | 35% |
| Interleukin 1 beta (IL-1 | 0.36 | 0 - 1.02 | 73% | 0.67 | 0.23 - 1.78 | 74% |
| Interleukin-6 (IL-6) | 0.89 | 0 - 3.02 | 59% | 1.05 | 0 - 3.48 | 57% |
| Interleukin-8 (IL-8) | 210.25 | 102.50 - 401.72 | 97% | 255.48 | 127.97 - 620.37 | 96% |
| Tumor necrosis factor-alpha (TNF | 1.06 | 0.49 - 2.26 | 59% | 0.70 | 0 - 2.02 | 58% |
|
Angiogenesis panel | ||||||
| Basic fibroblast growth factor (bFGF) | 0.65 | 0.0-1.70 | 61% | 0.76 | 0 - 1.58 | 60% |
| Fms Related Tyrosine Kinase 1 (FLT1) | 1758.01 | 1299.81 - 2353.27 | 99% | 1762.68 | 1318.11 - 2496.71 | 99% |
| Placenta growth factor (PLGF) | 67.29 | 37.04 - 118.33 | 99% | 72.96 | 39.13 - 144.45 | 99% |
| Vascular endothelial growth factor D (VEGF-D) | 293.81 | 229.89 - 449.74 | 99% | 274.81 | 188.84 - 411.39 | 98% |
|
Singleplex assays | ||||||
| Adiponectin | 21.15 | 15.15 - 28.46 | 100% | 20.26 | 14.43 - 26.27 | 100% |
| Epidermal growth factor (EGF) | 4373.99 | 2991.79 - 6053.52 | 100% | 4480.19 | 2881.58 - 5987.60 | 100% |
| Fas ligand (FASL) | 44.31 | 32.62 - 59.88 | 100% | 43.63 | 30.88 - 56.99 | 100% |
| Leptin | 605.21 | 0 - 1433.18 | 70% | 1047.06 | 334.97 - 2208.45** | 80% |
pg/mL for all analytes except adiponectin, which is μg/mL; analytes with <25% detectability excluded from table. If both duplicate levels were above the lower limit of detection (LLOD), the final calculated concentration was the mean of the two levels. If one technical replicate level was above the LLOD and one was below, then the final calculated concentration was the mean of the analyte above the LLOD and 1/2 the LLOD. If both analytes were below the LLOD then the result was recorded as zero.
Chi square = 4.96, p = 0.03
Associations of participant characteristics with analytes in human milk
Associations between analyte levels and select demographic and health status predictors adjusted for plate effects are shown in Table 3. Compared with White women, Black women had significantly higher levels of IL-1β (per category OR= 1.82, 95% CI = 1.19-2.79, p<0.001) and leptin (OR=2.56, 95% CI = 1.65-3.96, p<0.001). Obesity (BMI> 30 Kg/m2) was associated with significantly increased levels of INFγ, IL-1β, bFGF, EGF, FASL and leptin (all p-trend < 0.05). Obesity was also associated with statistically significantly lower levels of adiponectin (OR=0.56, 95% CI 0.32-0.97, p<0.001); however, the trend across all BMI categories did not reach statistical significance (p-trend = 0.06). Predictors including parity, baby’s age at donation, total lifetime days lactation and return of menses were related to several analytes as shown in Table S4, although values did not always show consistent trends across categories within variables.
Table 3.
Univariate odds ratios and 95% confidence intervals for associations between select maternal demographic predictors and analytes in human milk*
| Predictors | Race | Age (years: quartiles) | BMI (kg/m2: normal weight, overweight, obese) | |||||
|---|---|---|---|---|---|---|---|---|
| Analytes | Ref: White | Q1 19 - 28 Ref. | Q2 29 - 31 OR [95% CI] | Q3 32 - 35 OR [95% CI] | Q4 36 - 45 OR [95% CI] | ⩽24 Ref. | 25-<30 OR [95% CI] | ⩾30 OR [95% CI] |
| Pro-inflammatorv cytokines panel | ||||||||
| IFNγ | 1.10 [0.69, 1.68] | - | 0.86 [0.44, 1.67] | 1.22 [0.64. 2.31] | 1.04 [0.54, 1.99] | - | 1.68 [0.96, 2.93] | 2.03 [1.10, 3.74]*** |
| IL-1β | 1.82 [1.19, 2.79]** | - | 1.02 [0.57, 1.84] | 1.15 [0.65, 2.04] | 0.67 [0.37, 1.23] | - | 1.53 [0.94, 2.50] | 1.79 [1.05, 3.00]*** |
| IL-6 | 1.19 [0.77, 1.84] | - | 0.47 [0.25, 0.87] | 0.73 [0.40, 1.32] | 0.53 [0.29, 0.97] | - | 1.27 [0.76, 2.14] | 1.48 [0.85, 2.58] |
| IL-8 | 1.39 [0.91,2.15] | - | 0.82 [0.45, 1.46] | 1.14 [0.63,2.04] | 0.76 [0.42, 1.40] | - | 1.25 [0.76, 2.06] | 0.87 [0.50, 1.50] |
| TNFα | 0.66 [0.43, 1.01] | - | 0.69 [0.37, 1.26] | 1.20 [0.66, 2.17] | 0.18 [0.60, 1.95] | - | 1.75 [1.05, 2.91] | 1.46 [0.84, 2.53] |
| Anaioaenesis panel | ||||||||
| bFGF | 1.18 [0.78, 1.79] | - | 0.70 [0.40, 1.26] | 1.14 [0.64, 1.96] | 1.05 [0.59, 1.89] | - | 1.53 [0.94, 2.50] | 1.79 [1.06, 3.04]*** |
| FLT1 | 1.03 [0.68, 1.56] | - | 0.74 [0.41, 1.34] | 1.10 [0.63, 1.91] | 1.09 [0.61, 1.96] | - | 1.75 [1.08, 2.85] | 1.19 [0.70, 2.03] |
| PLGF | 1.31 [0.86, 1.98] | - | 0.64 [0.36, 1.42] | 1.12 [0.64, 1.94] | 1.02 [0.57, 1.83] | - | 1.31 [0.81,2.12] | 1.35 [0.79, 2.33] |
| VEGF-D | 0.72 [0.48, 1.10] | - | 0.99 [0.55, 1.79] | 0.85 [0.49, 1.50] | 0.86 [0.49, 1.51] | - | 1.54 [0.94, 2.52] | 0.75 [0.44, 1.29] |
| Sinaleplex | ||||||||
| Adiponectin | 0.70 [0.46, 1.07] | - | 1.05 [0.59, 1.88] | 0.89 [0.50, 2.58] | 1.16 [0.64, 2.11] | - | 0.95 [0.58, 1.56] | 0.56 [0.32, 0.97] |
| EGF | 0.92 [0.61, 1.41] | - | 1.48 [0.80, 2.71] | 1.41 [0.80, 2.49] | 1.26 [0.69, 2.30] | - | 2.34 [1.41, 3.87] | 1.51 [0.87, 2.61]*** |
| FASL | 1.06 [0.70, 1.61] | - | 0.81 [0.45, 1.47] | 1.20 [0.69, 2.12] | 0.97 [0.53, 1.76] | - | 1.92 [1.17, 3.15] | 2.05 [1.18, 3.57]*** |
| Leptin | 2.56 [1.65 - 3.96]** | - | 0.17 [0.70, 2.31] | 1.68 [0.94, 2.99] | 1.47 [0.81,2.63] | - | 5.83 [3.37, 10.18] | 19.88 [10.03,39.41]*** |
Note: Univariate analyses conducted using mixed effects ordinal logistic regression with random effect for plate, N= 292 women, with odds ratios (ORs) presented. Other predictors such as age at first birth, age at menarche, first-degree relative with breast cancer, and use of over the counter pain medicine were also tested, with null results.
p<0.05
p-trend<0.05
Multivariable models are shown in Table 4. All multivariable models included race; given the exploratory nature of this analysis, we additionally included covariates whose univariate association with the analyte of interest was p <0.4. As compared with White women, Black women had elevated levels of IL-1β and leptin, which persisted in multivariate models adjusted for BMI and other factors such as baby’s age in days. Overall, in multivariable models adjusted for race, elevated BMI was significantly associated with increased concentrations of IL-1β, bFGF, EGF, leptin and FASL (all p-trend < 0.05). Adiponectin was inversely associated with BMI, but this was not statistically significant. Return of menses since delivery was associated with increased levels of FLT1, PLGF and EGF. A history of smoking cigarettes was significantly positively associated with adiponectin concentrations, after controlling for race, BMI, and total lifetime days of lactation. Baby’s age was significantly associated with increased levels of IL-1β and IL-8 after controlling for race. Total lifetime days of lactation was generally associated with lower concentrations of bFGF and adiponectin, but the trends were not statistically significant (p-trends 0.56 and 0.70 respectively).
Table 4.
Multivariable odds ratios and 95% confidence intervals for associations of select participant characteristics with analytes in human milk*
| Predictor | IL-8 | IL-1β | bFGF | FLT1 | PLGF | EGF | Adiponectin | Leptin | FASL |
|---|---|---|---|---|---|---|---|---|---|
| OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | OR [95% CI] | |
| Race (ref. : White) | 1.20 [0.77-1.87] | 1.59 [1.03-2.45]** | 1.16 [0.75-1.79] | 0.93 [0.60-1.43] | 1.18 [0.77-1.79] | 0.85 [0.55-1.32] | 0.91 [0.58-1.44] | 2.16 [1.37-3.42]** | 0.99 [0.65-1.52] |
| Body mass index (BMI) | |||||||||
| <25 | N/A | ref | ref | ref | N/A | ref | ref | ref | ref |
| 25-<30 | 1.57 [0.94-2.60] | 1.51 [0.91-2.49] | 1.76 [1.06-2.90] | 2.31 [1.38-3.86] | 0.88 [0.53-1.41] | 6.01 [3.46-10.43] | 1.92 [1.17-3.15] | ||
| ⩾30 | 1.82 [1.05-3.16]*** | 1.77 [1.02-3.08]*** | 1.12 [0.65-1.95] | 1.49 [0.85-2.62]*** | 0.57 [0.32-1.01] | 18.36 [9.23-36.53]*** | 2.05 [1.17-3.58]*** | ||
| Menses since delivery | N/A | N/A | N/A | 2.37 [1.49-3.77]** | 2.84 [1.84-4.38]** | 1.75 [1.13-2.77]** | N/A | N/A | N/A |
| Pregnancies | |||||||||
| 1 | N/A | N/A | N/A | N/A | N/A | N/A | ref | N/A | N/A |
| 2 | 0.47 [0.26-0.85] | ||||||||
| 3+ | 0.63 [0.34-1.12] | ||||||||
| Baby’s age in days | |||||||||
| Q1 21 - 98 | ref | ref | ref | ref | N/A | N/A | N/A | N/A | N/A |
| Q2 100-189 | 2.21 [1.18-4.11] | 0.93 [0.51-1.72] | 0.36 [0.20-0.67] | 0.42 [0.23-0.77] | |||||
| Q3 194-315 | 4.15 [2.18-7.90] | 1.42 [0.77-2.62] | 0.47 [0.22-0.97] | 0.48 [0.26-0.89] | |||||
| Q4 316- 1500 | 8.74 [4.42-17.2*** | 2.44 [1.32-4.50]*** | 0.84 [0.43-1.68] | 0.77 [0.39-1.49] | |||||
| N/A | N/A | N/A | N/A | N/A | N/A | 1.82 [1.09-3.05]** | N/A | N/A | |
| Ever smoked cigarettes | |||||||||
| Total lifetime days of breastfeeding | |||||||||
| Q1 21 - 120 | N/A | N/A | ref | N/A | N/A | N/A | ref | N/A | N/A |
| Q2 127 - 255 | 0.47 [0.22-1.02] | 0.35 [0.19-0.64] | |||||||
| Q3 270 - 735 | 0.56 [0.29-1.08] | 0.73 [0.38-1.38] | |||||||
| Q4 750 - 3900 | 0.61 [0.31-1.21] | 0.87 [0.42-1.40] | |||||||
Associations tested in ordinal logistic regression with random effect for plate. Multivariable models were built in forward stepwise selection, with race and all predictors p< 0.4 retained in final models. Factors with cells labeled N/A (Not Applicable) were not included in the model for the analyte listed because they failed to meet criteria for selection.
p<0.05
p-trend<0.05
Exploratory factor analysis of analytes identified two factors with significant contributions, with positive loading of VEGFD, FLT1, FASL, EGF and PLGF to Factor 1, and of IL-6, IL-1β, and TNFα to Factor 2 (Table 5). Factor 1 was predominantly comprised of angiogenesis factors, but also FASL, which functions in apoptosis, whereas significant loading on Factor 2 primarily included pro-inflammatory cytokines. In univariate analysis (Table 6), higher levels of Factor 1 scores were associated with increased age, BMI and with resumption of menses since delivery (all p < 0.05). Higher levels of Factor 2 scores were inversely associated with baby’s age (p-trend < 0.05) and positively with mother’s age (p-trend < 0.01). Multivariable analyses for factor scores, which included race in all of the models, demonstrated significant positive relationship of Factor 1 scores with BMI and menses since delivery (p < 0.05), controlling for maternal age and baby’s age. The multivariable model for Factor 2 showed that higher scores were associated with Black race (p < 0.05), but had an inverse association with baby’s age in days (p-trend <0.01), controlling for maternal age and race (Table 7).
Table 5.
Factor patterns - varimax rotation of analyte factor correlations of residuals after removal of plate effects.
| Factor 1 | Factor 2 | |
|---|---|---|
| VEGFD | 84 * | 0 |
| FLT1 | 84 * | 3 |
| FASL | 73 * | 3 |
| EGF | 70 * | 9 |
| PLGF | 69 * | 9 |
| Leptin | 47 | 11 |
| Adiponectin | 47 | 31 |
| BFGF | 46 | 37 |
| IL-8 | 38 | 10 |
| IL-6 | 10 | 93 * |
| IL-1β | 7 | 79 * |
| TNFα | 13 | 74 * |
| IFNγ | 8 | 26 |
Values are multiplied by 100 and rounded to the nearest integer. Starred (*) values represent greater than 49% of the variance associated with a factor.
Table 6.
Univariate ordinal logistic regression of demographic predictors on quartiles of factor scores*
| Predictor | Factor 1-Growth | Factor 2-Inflammation | ||
|---|---|---|---|---|
| OR | 95% CI | OR | 95% CI | |
| Race | 1.28 | [0.84 - 1.93] | 1.43 | [0.94 - 2.16] |
| Age | ||||
| Q1 19-28 | ref | ref | ||
| Q2 29- 31 | 1.09 | [0.61 - 1.92] | 0.75 | [0.42 - 1.33] |
| Q3 32 - 35 | 1.61 | [0.92 - 2.80] | 1.12 | [0.63- 1.97] |
| Q4 36 - 45 | 1.71 | [0.95 - 3.10]*** | 1.01 | [0.56 - 1.79]*** |
| Body mass index (BMI) | ||||
| <25 | ref | ref | ||
| 25 - <30 | 2.13 | [1.31-3.46] | 1.24 | [0.76-2.01] |
| ⩾30 | 1.92 | [1.11-3.20]** | 1.25 | [0.73-2.14] |
| Baby’s age in days | ||||
| Q1 21-98 | ref | ref | ||
| Q2 100 - 89 | 0.72 | [0.41-1.26] | 0.50 | [0.28-0.90] |
| Q3 194-315 | 0.93 | [0.52- 1.65] | 0.39 | [0.22-0.71] |
| Q4 316 - 1500 | 1.48 | [0.82-2.67] | 0.45 | [0.25-0.82]** |
| Menses since delivery | 2.07 | [1.35 - 3.17]** | 0.89 | [0.59 - 1.36] |
| Ever smoked cigarettes | 1.15 | [0.72 - 1.83] | 1.04 | [0.66 - 1.65] |
| Pregnancies 1 | ref | ref | ||
| 2 | 1.13 | [0.68, 1.90] | 0.61 | [0.36, 1.04] |
| 3+ | 1.12 | [0.69, 1.81] | 0.77 | [0.48, 1.25] |
| Lifetime days of breastfeeding | 1.12 | [0.93, 1.351 | 0.86 | [0.72, 1.041 |
Associations tested in mixed effects multiple level ordinal logistic regression on quartiles of factor scores with a random effect for plate. Factors with cells labeled N/A (Not Applicable) were not included in the model for the analyte listed because they failed to meet criteria for selection.
p<0.05
p-trend<0.05
Table 7.
Multivariable ordinal logistic regression of demographic predictors on quartiles of factor scores*
| Predictor | Factor 1 - Growth | Factor 2 - Inflammation | ||
|---|---|---|---|---|
| OR | 95% CI | OR | 95% CI | |
| Race | 1.11 | [0.74 - 1.74] | 1.59 | [1.03 - 2.39]** |
| Age | ||||
| Q1 19-28 | ref | ref | ||
| Q2 29 - 31 | 1.09 | [0.61 - 1.96] | 0.84 | [0.47 - 1.50] |
| Q3 32 - 35 | 1.42 | [0.80-2.51] | 1.25 | [0.70-2.23] |
| Q4 36 - 45 | 1.62 | [0.88-2.95] | 1.14 | [0.64, 2.06] |
| Body mass index (BMI) | ||||
| <25 | ref | N/A | ||
| 25 - <30 | 2.04 | [1.24-3.34] | ||
| ⩾30 | 1.61 | [0.93-2.81]** | ||
| Baby’s age in days | ||||
| Q1 21-98 | ref | ref | ||
| Q2 100 - 189 | 0.72 | [0.41 - 1.26] | 0.50 | [0.28-0.90] |
| Q3 194 - 315 | 0.93 | [0.52 - 1.65] | 0.38 | [0.21-0.69] |
| Q4 316 - 1500 | 1.48 | [0.82-2.67] | 0.45 | [0.22-0.75]*** |
| Menses since delivery | 1.98 | [1.29-3.05]** | N/A | |
Associations tested in ordinal logistic regression on quartiles of factor scores with a random effect for plate. Multivariable models were built in forward stepwise selection with race and all predictors p< 0.4 retained in final models. Factors with cells labeled N/A (Not Applicable) were not included in the model for the analyte listed because they failed to meet criteria for selection.
p<0.05
p-trend<0.05
Discussion
Our exploratory analysis demonstrates the feasibility of measuring many analytes in human milk in epidemiologic studies seeking to identify differences in breast biology between Black and White women. Preliminarily, we find that certain analytes in human milk may vary by BMI, race, and other characteristics of the lactating mother and nursing baby. Thus, we suggest that analysis of human milk may inform hypotheses about early events in breast carcinogenesis for women with different levels of early-onset breast cancer risk.
We successfully analyzed 13 out of 15 candidate analytes tested but excluded TIE2 and VEGFC because of poor detectability. The other angiogenic and adipokines markers achieved reasonable CVs and ICCs; for the cytokines, the CVs were adequate given the relatively high ICCs (for the most part exceeding 90%). Agarwal’s 2011 review (22) of 25 pro-inflammatory cytokine studies of human milk included some of the analytes measured in our study with some differences and similarities in results. Among reviewed reports, levels of IFNγ, IL-6, and TNF-α were higher than in our study, though their ranges fell within the wide range of median quantities reported for each analyte, whereas levels of IL1-β and of IL-8 were similar across studies (22). Most summarized studies used singleplex enzyme-linked immunosorbent assays (ELISA), while our study used electrochemoluminescent, multiplex technology. In addition, another systematic review of 26 studies of leptin concentrations in milk, primarily measured with ELISA, found levels similar to our data (18).
Our results suggest that BMI and race represent potential correlates of several analytes in human milk. Although highly preliminary, our work demonstrates that levels of leptin in human milk are positively associated with BMI, as expected and previously reported (reviewed in 18 and 19-21), and that levels are higher among Black as compared with White women, even after controlling for BMI, findings which are consistent with prior studies (23, 24). Many findings implicate leptin in breast carcinogenesis, including studies of obese women showing higher ratios of circulating leptin to BMI in breast cancer patients versus age-matched controls, and data suggesting that leptin is increased in breast cancer tissues compared with benign specimens (24, 25). Epidemiologic studies of circulating leptin and breast cancer risk are conflicting, but inverse associations with premenopausal breast cancer have been reported. (26, 27) In our analysis, obesity was also associated with higher milk levels of IL-1β, bFGF, EGF, leptin and FASL. A prior analysis of serial milk samples from normal-weight and heavy women found associations between higher BMI and elevated leptin and insulin levels, with substantially higher concentrations of insulin in breastmilk versus plasma (28). Similar to our findings, BMI was not significantly related to levels of IL-6, IL-8 and TNFα (28).
Our data suggest that milk from Black women may contain higher levels of IL-1β than in White women, independent of BMI and lactation duration. IL-1β is an important “alarm” inflammatory mediator produced by monocytes and macrophages in response to cellular necrosis that has important effects in multiple aspects of tumorigenesis (29). This finding contrasts with that of Burch et al. (30), who measured associations between demographic characteristics, allergy history and immune factors in human milk in a group of 115 women of different racial groups: 77 White, 20 Black, and 18 “Other”. They noted that breast milk of Black women had higher levels of IL-6 (6.60 vs. 1.55 pg/mL) and IFN-γ (6.56 vs. 2.58 pg/mL) than that of White women, controlling for covariates such as history of atopy, smoking, and BMI. There was no statistically significant difference in IL1-β levels in Black and White women in the study, perhaps due to low statistical power, lack of samples from late in the breastfeeding trajectory, or different control variables.
We and others have hypothesized that increased inflammation within the breast of Black women compared with White women might be linked to the higher rates of early onset breast cancer in Black women (31, 32). The transformation of the lactating breast back to an inactive gland (involution) has long-lasting effects on the health of the breast, as evidenced by a study in mice that found evidence that involution is similar to the tissue remodeling that occurs during wound healing (33). Thus, our preliminary data showing that Black race and obesity arelinked to higher levels of inflammatory cytokines and growth factors in human milk, if confirmed, suggest potential strategies to reduce racial disparities in breast cancer risk through interventions such as weight control and perhaps short courses of anti-inflammatory agents. This would support the preclinical results reported by Lyons et al. showing that a postpartum pro-inflammatory mechanism may promote development of aggressive breast cancer and that this process can be inhibited by NSAIDs (33).
Along with the poor detectability and reliability of some analytes in human milk, challenges in this study included the heterogenous nature of milk collection (local vs. shipped samples). Specifically, nearly all specimens donated by Black women were shipped, whereas specimens from White women were collected near the laboratory and more rapidly processed, raising concerns that mode of collection could confound associations between participant characteristics and analytes levels. Although collection of milk samples by mail offers reduced control over sample handling, it offers advantages with respect to enrolling large numbers of diverse participants. Still, the difference in collection methods represents a limitation that we attempted to minimize through statistical adjustment. Our findings are supported by a classic review by Lawrence (34) that concluded that milk proteins and immune components are relatively stable over a range of collection and storage conditions, and another study showed that leptin levels did not vary between fresh, frozen and pasteurized milk samples (35). In our study, all women donated specimens using their own breast pump, following specific instructions for standardized collection. Table 2 demonstrates that shipped samples have higher levels of leptin, while all other analytes had similar levels, suggesting that location of donation did not have major influences on most cytokines. When we stratified analyses by shipping, leptin was positively associated with BMI in both shipped and locally collected samples (data not shown); therefore, we concluded that differences in leptin levels by collection method were likely due to race, not method of collection, although this interpretation is tentative.
Strengths of this analysis include collection of a large number of samples from a racially diverse group of women with comprehensive annotation, and the use of standardized collection methods. The innovative use of local recruitment along with the Internet to facilitate an off-site collection protocol, allowed for representation of lactating women throughout the lactation trajectory. In our cross-sectional study, we treated the length of lactation largely as a control variable in analysis. Other longitudinal studies demonstrated changes in cytokine expression related to length of lactation (13, 36, 37), though with different cytokine panels and different durations of lactation, limiting our ability to integrate these findings with ours. For example, Qin et al. (13) recruited women who provided “transitional”, “mature” and “wean” milk at varying timepoints, using breastfeeding pattern to differentiate the samples instead of number of days since birth of the baby. Chollet-Hinton et al. (36) and Collado et al. (37) analyzed milk from women at varying timepoints up to 12 weeks in the breastfeeding trajectory, with no indication that any sample was a wean sample. Future research in this area with collection of repeated samples over time from larger numbers of women, with comprehensive annotation of feeding patterns, medical history, and medication use throughout the postpartum period, would represent an important next step. Additional specimens could include both serum and tissue samples from matched time points in the breastfeeding trajectory.
The ease of collection of human milk, along with the potential to assess biomarkers such as adipokines, cytokines, and growth factors, as well as genetic and epigenetic changes in luminal epithelial cells (15), suggest that human milk analysis may inform our understanding of breast carcinogenesis. If high levels of inflammation in human milk are linked to breast cancer risk, women with such forms of “inflammatory mastopathy” might be candidates for trials assessing screening at early ages, and possibly short-term chemoprevention with anti-inflammatory agents. While this hypothesis is supported by data from elegant preclinical models (33), testing this hypothesis among women will require both methods development and large population-based studies.
Supplementary Material
Table S1. Demographic and health characteristics of participants stratified by collection type
Table S2. Quality control: repeated samples across 8 plates*
Tables S3. Coefficients of variation (CV) and intraclass correlation coefficients (ICC) from technical duplicate measures of analytes in breastmilk from 292 women
Table S4. Univariate odds ratios (OR) and 95% confidence intervals (CI) between maternal predictors and analytes, in quartiles, tertiles or above/below median, using mixed effects ordinal logistic regression with random effect for plate, N = 292 women
Footnotes
Conflict of interest statement: The authors state they have no conflict of interest.
References
- 1.Anderson KN, Schwab RB, Martinez ME. Reproductive risk factors and breast cancer subtypes: a review of the literature. Breast Cancer Res Treat. 2014. 144(1):1–10. DOI: 10.1007/s10549-014-2852-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Holm J, Eriksson L, Ploner A, et al. Assessment of breast cancer risk factors reveals subtype heterogeneity. Cancer Research. 2017. 2017. July 1;77(13):3708–3717. DOI: 10.1158/0008-5472.CAN-16-2574 [DOI] [PubMed] [Google Scholar]
- 3.Palmer JR, Viscidi E, Troester MA, et al. Parity, lactation and breast cancer subtypes in African American women: Results from the AMBER consortium. J Natl Cancer Inst. 2014;106(10):l–8. DOI: 10.1093/jnci/dju237 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Victora CG, Bahl R, Barros AJD, et al. Breastfeeding in the 21st century: epidemiology, mechanisms, and lifelong effect. Lancet. 2016; 387:475–490. DOI: 10.1016/S0140-6736(15)01024-7 [DOI] [PubMed] [Google Scholar]
- 5.DeSantis C, Ma J, Sauer AG, et al. Breast Cancer Statistics, 2017, Racial Disparity in Mortality by State. CA Cancer J Clin. 2017; DOI: 10.3322/caac.2141 [DOI] [PubMed] [Google Scholar]
- 6.Anstey EH, Chen J, Elam-Evans LD, Perrine CG. Racial and Geographic Differences in Breastfeeding — United States, 2011-2015. MMWR Morb Mortal Wkly Rep 2017;66:723–727. DOI: 10.15585/mmwr.mm6627a3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Flegel KM, Kruszon-Moran D, Carroll MD. Trends in Obesity Among Adults in the United States, 2005 to 2014. JAMA. 2016;315(21):2284–2291. DOI: 10.1001/jama.2016.6458. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Pierobon M, Frankenfeld CL. Obesity as a risk factor for triple-negative breast cancers: a systematic review and meta-analysis. Breast Cancer Res Treat. 2013. January;137(1):307–14. DOI: 10.1007/s10549-012-2339-3. [DOI] [PubMed] [Google Scholar]
- 9.Hawkes JS, Bryan DL, Gibson RA (2002) Cytokine production by human milk cells and peripheral blood mononuclear cells from the same mothers. J Clin Immunol 22(6):338–344. [DOI] [PubMed] [Google Scholar]
- 10.Hines EP, Rayner JL, Barbee R, Moreland RA, Valcour A, Schmid JE, Fenton SE (2007) Assays for endogenous components of human milk: comparison of fresh and frozen samples and corresponding analytes in serum. J Hum Lact 23(2): 144–156. DOI: 10.1177/0890334407300334 [DOI] [PubMed] [Google Scholar]
- 11.Ozarda Y, Gunes Y, Tuncer GO (2012) The concentration of adiponectin in breast milk is related to maternal hormonal and inflammatory status during 6 months of lactation. Clin Chem Lab Med 50(5):911–917. DOI: 10.1515/cclm-2011-0724 [DOI] [PubMed] [Google Scholar]
- 12.Arcaro KF, Browne EP, Qin W, et al. Differential expression of cancer-related proteins in paired breastmilk samples from women with breast cancer. J Hum Lact. 2012;28:543–47. DOI: 10.1177/0890334412453205 [DOI] [PubMed] [Google Scholar]
- 13.Qin W, Zhang K, Kliethermes B, et al. Differential expression of cancer associated proteins in breastmilk based on age at first full term pregnancy. BMC Cancer. 2012. March 21;12:100 DOI: 10.1186/1471-2407-12-100 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Yang HP, Schneider SS, Chisholm CM, et al. Association of TGF-β2 levels in breastmilk with severity of breast biopsy diagnosis. Cancer Causes Control. 2015. March;26(3):345–54. DOI: 10.1007/s10552-014-0498-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Murphy J, Sherman ME, Browne EP, et al. Potential of breastmilk analysis to inform early events in breast carcinogenesis: rationale and considerations. Breast cancer research and treatment. 2016. May;157(1):13–22. DOI: 10.1007/s10549-016-3796-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wong CM, Anderton DL, Smith-Schneider S, Wing MA, Griven MC, Arcaro K. Quantitative analysis of promoter methylation in exfoliated epithelial cells isolated from breast milk of healthy women. Epigenetics. 2010. Oct 1;5(7):645–55. DOI: 10.4161/epi.5.7.1296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Browne EP, Punska EC, Lenington S, Otis CN, Anderton DL & Arcaro KF. Increased promoter methylation in exfoliated breast epithelial cells in women with a previous breast biopsy. Epigenetics. 2011;6:1425–35. DOI: 10.4161/epi.6.12.18280 [DOI] [PubMed] [Google Scholar]
- 18.Andreas NJ, Hyde MJ, Gale C, et al. Effect of maternal body mass index on hormones in breast milk: a systematic review. PloS one. 2014;9(12):e11 DOI: 10.1371/journal.pone.0115043 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Savino F, Sardo A, Rossi L, Benetti S, Savino A, Silvestro L. Mother and Infant Body Mass Index, Breast Milk Leptin and Their Serum Leptin Values. Nutrients. 2016;8(6). DOI: 10.3390/nu8060383. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Weyermann M, Beermann C, Brenner H, Rothenbacher D. Adiponectin and leptin in maternal serum, cord blood,and breast milk. Clinical chemistry. 2006;52(11):2095–2102. DOI: 10.1373/clinchem.2006.071019 [DOI] [PubMed] [Google Scholar]
- 21.Andreas NJ, Hyde MJ, Herbert BR, et al. Impact of maternal BMI and sampling strategy on the concentration of leptin, insulin, ghrelin and resistin in breast milk across a single feed: a longitudinal cohort study. BMJ open. 2016;6(7):e010778 DOI: 10.1136/bmjopen-2015-010778 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Agarwal S, Karmaus W, Davis S, Gangur V. Immune markers in breast milk and fetal and maternal body fluids: a systematic review of perinatal concentrations. Journal of Human Lactation. 2011;27(2): 171–186. DOI: 10.1177/0890334410395761 [DOI] [PubMed] [Google Scholar]
- 23.Cohen SS, Fowke JH, Cai Q, et al. Differences in the association between serum leptin levels and body mass index in black and white women: a report from the Southern Community Cohort Study. Annals of nutrition & metabolism. 2012;60(2):90–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Azrad M, Gower BA, Hunter GR et al. Racial differences in adiponectin and leptin in healthy premenopausal women. Endocrine (2013) 43: 586 DOI: 10.1007/s12020-012-9797-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Barone I, Giordano C, Bonofiglio D, Andò S, Catalano S. Leptin, obesity and breast cancer: progress to understanding the molecular connections. Current Opinion in Pharmacology. 2016:31;83–89. doi: 10.1016/j.coph.2016.10.003 [DOI] [PubMed] [Google Scholar]
- 26.Matthews SB, Thompson HJ. The Obesity-Breast Cancer Conundrum: An Analysis of the Issues. Marchetti D, ed. International Journal of Molecular Sciences. 2016;17(6):989. doi: 10.3390/ijms17060989. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Harris HR, Tworoger Shelley S.,Hankinson SE, Rosner BA, Michels KB. Plasma leptin levels and risk of breast cancer in premenopausal women. Cancer prevention research (Philadelphia, Pa). 2011;4(9):1449–1456. doi: 10.1158/1940-6207.CAPR-11-0125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Young B, Patinkin Z, Palmer C, de La Houssaye B, Barbour LA, Hernandez T, Friedman, J, Krebs, NF. Human milk insulin is related to maternal plasma insulin and BMI - But other components of human milk do not differ by BMI. Eur J Clin Nutr. 2017. September; 71(9): 1094–1100. DOI: 10.1038/ejcn.2017.75 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Voronov E, Dotan S, Krelin Y, Song X, Elkabets M, Carmi Y, Rider P, Cohen I, Romzova M, Kaplanov I, Apte R. Unique versus redundant functions of IL-1α and IL-1β in the tumor microenvironment. Frontiers in Immunology. 2013;4:177 DOI: 10.3389/fimmu.2013.00177 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Burch J, Karmaus W, Gangur V, Soto-Ramírez N, Yousefi M, Goetzl LM. Pre- and perinatal characteristics and breast milk immune markers. Pediatr Res. 2013. November;74(5):615–21. doi: 10.1038/pr.2013.141. [DOI] [PubMed] [Google Scholar]
- 31.Fornetti J, Martinson HA, Betts CB, et al. Mammary gland involution as an immunotherapeutic target for postpartum breast cancer. J Mammary Gland Biol Neoplasia. 2014;2:213–28. DOI: 10.1007/s10911-014-9322-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Faupel-Badger JM, Arcaro KF, Balkam JJ, et al. Postpartum remodeling, lactation, and breast cancer risk: summary of a National Cancer Institute-sponsored workshop. Journal of the National Cancer Institute. 2013;105(3):166–174. DOI: 10.1093/jnci/djs505 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Lyons TR, O’Brien J, Borges VF, et al. Postpartum mammary gland involution drives progression of ductal carcinoma in situ through collagen and COX-2. Nature medicine. 2011;17(9): 1109–1115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Lawrence RA. Storage of human milk and the influence of procedures on immunological components of human milk. Acta paediatrica. Supplement. 1999;88(430):14–18. [DOI] [PubMed] [Google Scholar]
- 35.Chang JC, Chen CH, Fang LJ, Tsai CR, Chang YC, Wang TM. Influence of prolonged storage process, pasteurization, and heat treatment on biologically-active human milk proteins. Pediatrics and neonatology. 2013;54(6):360–366. DOI: 10.1016/j.pedneo.2013.03.018 [DOI] [PubMed] [Google Scholar]
- 36.Chollet-Hinton LS, Stuebe AM, Casbas-Hernandez P, Chetwynd E, Troester MA. Temporal trends in the inflammatory cytokine profile of human breastmilk. Breastfeed Med. 2014. December;9(10):530–7. DOI: 10.1089/bfm.2014.0043. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Collado MC, Santaella M, Mira-Pascual L, Martinez-Arias E, Khodayar-Pardo P, Ros G, Martinez-Costa C. Longitudinal study of cytokine expression, lipid profile and neuronal growth factors in human breast milk from term and preterm deliveries. Nutrients. 2015. October 19;7(10):8577–91. doi: 10.3390/nu7105415.</References> [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. Demographic and health characteristics of participants stratified by collection type
Table S2. Quality control: repeated samples across 8 plates*
Tables S3. Coefficients of variation (CV) and intraclass correlation coefficients (ICC) from technical duplicate measures of analytes in breastmilk from 292 women
Table S4. Univariate odds ratios (OR) and 95% confidence intervals (CI) between maternal predictors and analytes, in quartiles, tertiles or above/below median, using mixed effects ordinal logistic regression with random effect for plate, N = 292 women