Hormonal determinants of mammographic density and density change

In this nested study within the large KARMA cohort, we included 1040 clinically healthy controls without any prior breast cancer diagnosis or other cancer and who were not using MHT at the time of blood draw. No women with previous gynaecological surgery were included. The samples were previously randomly selected as age-matched controls to breast cancer cases within the KARMA cohort. KARMA is a population-based prospective cohort study initiated in 2011 comprising 70,877 women attending mammography screening or clinical mammography in Sweden [31, 32]. The overarching goal of KARMA is to reduce the incidence and mortality of breast cancer by focusing on individualised prevention and screening.

Women completed a comprehensive KARMA baseline questionnaire and donated non-fasting EDTA plasma samples of peripheral blood at enrolment [31, 32]. All variables included in the analyses were collected using the web-based questionnaire at study entry. Baseline BMI was self-reported.

Each study participant signed a written informed consent form and accepted linkage to national breast cancer registers. The Stockholm ethical review board approved the study (2010/958-31/1).

Processed mammograms from mediolateral oblique and craniocaudal views of the left and right breasts were collected from full-field digital mammography system at study enrolment [31, 32]. We used average dense area (cm²) (over the left and right breasts) using the STRATUS area-based method. STRATUS is a fully automated tool developed to analyse digital and analogue images using an algorithm that measures density on all types of images regardless of vendor [33]. When studying repeated mammograms from the same individual women, it is important to consider the technical differences between mammogram. In the current study, mammograms from the same women were aligned before density measures were performed. The concept of the alignment method has been described previously [33], as has the calculation of MD change over time [15].

All blood samples were collected at study entry and handled in accordance with a strict 30-h cold-chain protocol at the Karolinska Institutet high-throughput biobank. Hormones were measured in blinded peripheral blood plasma by the UPSFC-MS/MS system (Waters Corporation, Milford, USA), as described previously [30]. Sample preparation for the analysis of desulfated steroid hormones was carried out through liquid-liquid extraction with tert-butyl methyl ether (MTBE) followed by derivatisation with methoxyamine. Sulphated DHEA (DHEAS) was analysed directly, after extraction with MTBE after protein precipitation. The separation of the desulfated steroids and DHEAS was accomplished using the Acquity UPC² BEH and CSH fluoro-phenyl columns (3.0 mm × 150 mm, 1.7 μm), respectively (Waters, Milford, USA). Desulfated steroid methoxyamine derivatives were separated using 0.1% formic acid in methanol-isopropanol (1:1, v/v) (2 mL/min) as a modifier whereas DHEAS was separated using 10 mM ammonium acetate in methanol with 3% (v/v) water (1.5 mL/min) using the respective columns. The MS detection was performed using electrospray ionisation in the positive ionisation mode ESI+ve for desulfated steroid methoxyamine and positive ionisation mode ESI-ve for DHEAS derivatives, with nitrogen as desolvation gas and argon collision gas. Data acquisition range was set to m/z 100–600. The quantification was based on a multiple reaction monitoring (MRM) method; collision energy and cone voltage were set as described previously [30], using individual analysis of standard desulfated steroids and DHEAS (100 ng/mL). Quantification of hormones was performed using the suitable deuterated internal standards, and limit of quantification (LOQ) of desulfated steroids (0.05–0.5 ng/mL) and DHEAS (0.01 ng/mL). The coefficient of variation of desulfated steroids and DHEAS assays was < 7.2% and < 3.2%, respectively. Limit of detection (LOD) and LOQ were determined as the lowest concentration which provided a signal-to-noise ratio (S/N) greater than 3 and 10, respectively, by repeated injection (n = 6) with a relative standard deviation of replicates below 15%. Values were missing for all hormones and across all batches. Information on LOQ and linear range for all steroinds can be found in [30]. Data was acquired, analysed and processed using the MassLynx TM4.1 software (Waters, Milford, USA).

The peptide hormones SHBG (cat.no. DSHBG0B) and prolactin (cat.no. DPRL00) were measured by using immuno-assay kits from R&D Systems (Minneapolis, USA). Both sandwich-type assays used a pre-coated 96-well plate and a supply of enzyme-labelled secondary antibody, and standards, and were analysed according to the manufacturer’s instructions. The resulting absorbance was read in a BioRad 680 Microplate Reader (BioRad Laboratories, Hercules, CA) at 450 nm with 595 nm as background. The goodness of fit was verified by the r² values. The LOQ was 2.0 nM for SHBG and 0.6 ng/mL for prolactin.

Multivariable adjusted linear regression models were used to estimate the association of endogenous hormones with baseline mean MD and 95% confidence interval (CI), as well as annual MD change (95% CI). Annual MD change over the follow-up period was estimated for each woman as a slope using a linear regression on age at each MD measurement. We calculated the geometric mean of baseline MD or MD change within tertile distributions for each hormone, and the difference between each category and the reference by multivariable-adjusted analyses of variance. P for trend was calculated by linear regression with baseline MD or MD change as a dependent continuous variable across tertiles of hormones. All models were adjusted for age and BMI at baseline, time of day of blood draw and plasma sample plate number to account for missingness by technical error. All models for MD change were additionally adjusted for physical activity (MET-h/d) at baseline. Hormones were natural log-transformed. Linear regression models with continuous variables to estimate the association of endogenous hormones with baseline mean MD and annual MD change were also stratified by menopausal status defined at baseline. All P values were two-sided and considered statistically significant if < 0.05. Analyses were conducted using SPSS (version 26; IBM Corporation).

Baseline characteristics of the 1040 women included in the study are presented in (Table 1). The average age of participants at study entry and mammography was 57.9 years (SD 9.3). Three hundred thirty-five women were premenopausal (mean age 46.8, SD 3.9) and 705 were postmenopausal (mean age 63.1 and SD 5.9), at study entry. The average MD area was 24.9 cm² (SD 22.2) (premenopausal 37.4, SD 24.7; postmenopausal 19.0 SD 18.2), and the average MD change was − 0.8 cm² of dense area per year (SD 3.3) (premenopausal − 1.5, SD 4.3; postmenopausal − 0.5, SD 2.7). On average, 2.8 (median 3.0) mammography screening examinations were available to calculate the annual MD change. The average time spread for the follow-up mammograms were between 12 and 24 months in the study.

We measured 17 hormones from the progestogen (n = 3), androgen (n = 7), oestrogen (n = 2) and corticoid (n = 5) pathways, as well as peptide hormones and proteins (n = 2) (Table 1). Concentrations of pregnenolone, progesterone, 17OH-progesterone, DHEA, DHEAS, androstenedione, oestrone sulphate and prolactin (all P < 0.001); androsterone (P = 0.001); and etiocholanolone (P = 0.023) were all significantly lower in postmenopausal compared to premenopausal women. The overall range of quantification was between 43.3 and 99.5% (Table 1). Missing values were technical and not associated with menopausal status.

The influence of endogenous hormone concentrations on MD (cm²) in the entire population is shown in Table 2 and Fig. 1. A doubling of progesterone concentration corresponded to an increase of + 1.29 cm² in baseline MD (P < 0.001). Similar was seen for 17OH-progesterone (+ 1.09 cm²; P = 0.028), oesterone sulphate (+ 1.42 cm²; P = 0.034), prolactin (+ 2.11 cm²; P = 0.049) and SHBG (+ 4.18 cm²; P < 0.001). Women in the highest tertile (Q3) of progesterone had an average baseline MD of 29.26 cm² as compared with 24.39 cm² for women in the lowest tertile (Q1) (P_difference = 0.014) (Fig. 1; Additional file 1: Table S1). Similar was seen for 17OH-progesterone (Q3: 28.19 cm² versus Q1: 23.71 cm²; P_difference = 0.007) and SHBG (Q3: 28.02 cm² versus Q1: 22.52 cm²; P_difference = 0.001).

A higher concentration of DHEA and 11-deoxycortisol was inversely associated with baseline MD. Women in Q3 of DHEA had an average baseline MD of 23.59 cm² versus 28.65 cm² in Q1 (P_difference = 0.029), corresponding to a lower baseline MD of − 1.33 cm² per doubling of hormone (P = 0.033). Women in Q3 of 11-deoxycortisol had an average baseline MD of 23.68 cm² versus 27.28 cm² in Q1 (P_difference = 0.036), corresponding to a lower baseline MD of − 1.33 cm² per doubling of hormone (P = 0.033).

When stratifying by menopausal status, progesterone was positively associated with baseline MD among premenopausal (+ 1.78 cm² per doubling of hormone; P = 0.004) (Table 3), but not postmenopausal (− 0.07 cm²; P = 0.888), women (Table 4). SHBG was positively associated with baseline MD in both premenopausal and postmenopausal women (+ 6.58 cm², P = 0.004; + 2.60 cm², P = 0.042, respectively). Other hormones did not reach statistical significance in the stratified analyses.

The influence of endogenous hormone concentrations on MD change (cm²/year) in the entire population is shown in Table 2. Greater concentrations of hormones were associated with MD change for hormones in the androgen pathway. Androstenedione was associated with lesser MD change (0.34 cm²/year per doubling of hormone concentration; P = 0.005), as was testosterone (0.35 cm²/year; P = 0.004) and free testosterone (0.31 cm²/year; P = 0.004) (Table 2). Women in Q3 of testosterone had an average MD change decrease of − 0.38 cm²/year versus − 1.11 cm²/year for women in Q1 (P_difference = 0.019) (Fig. 2 and Additional file 1: Table S2). Similarly, women in Q3 of free testosterone had an average MD change of − 0.48 cm²/year versus − 1.14 cm²/year for women in Q1 (P_difference = 0.032). Women in Q3 of androstenedione had an average MD change of − 0.48 cm²/year versus − 1.05 cm²/year women in Q1; however, the difference was not statistically significant (P_difference = 0.076).

When stratifying by menopausal status, only free testosterone was significantly associated with MD change (0.53 cm²/year per doubling of hormone concentration; P = 0.030) among premenopausal women (Table 3). Total testosterone was borderline significant associated with MD change (P = 0.057). In postmenopausal women, MD change was significantly associated with DHEA (0.34 cm²/year; P = 0.001), androstenedione (0.30 cm²/year; P = 0.008) and androsterone (0.27 cm²/year; P = 0.006) (Table 4). Total testosterone was borderline significant associated with MD change in postmenopausal women (P = 0.069).