Key words
Vitamin D - breast density - BMI - age - menopausal status
Introduction
In recent years there has been a growing public focus on vitamin D and its effect
on processes in the human body; the number of publications on this topic continues
to grow [1]. Vitamin D is not just responsible for regulating calcium levels, it also has a
wide range of immunological and anti-proliferative effects. Because of these additional
pleiotropic functions, the potential role of vitamin D in reducing the risk for various
epithelial cancers including the risk for breast cancer is currently being discussed
[2].
Vitamin D is a prohormone which can be ingested enterally but is predominantly produced
by the body itself with the help of sunlight. The term “vitamin” is therefore wrong
in the proper sense of the word, as true vitamins are exclusively supplied through
food intake. The first step in producing vitamin D occurs in the liver where cholesterol
is converted into 7-DHC (7-dehydrocholesterol) [3]. After transportation to the skin it is converted into vitamin D3 (cholecalciferol) with the help of UVB radiation [4]. Cutaneous vitamin D3 synthesis increases exponentially according to the amount of sunlight in the sky,
and in the northern hemisphere it reaches its maximum during the summer months [5]. During this period and depending on the time spent outdoors, skin type, and location,
the skin is the main supplier of vitamin D in our body, with up to 90 percent of vitamin
D created by dermal synthesis. The intake of foodstuffs containing vitamin D3 such as fish, mushrooms and dairy products also contributes to the provision of vitamin
D3; particularly in winter this can be important for vitamin D metabolism.
Ingested or synthesized vitamin D3 is hydroxylated in the liver to 25-hydroxyvitamin D [25(OH)D, calcidiol} [6], [7]. Calcidiol is then transported to the kidneys and other tissues mainly via the bloodstream
where it is converted into its biologically active metabolite, 1,25-dihydroxyvitamin
D (calcitriol) [6], [8]. Calcitriol binds to intracellular vitamin D receptors and through complex formation
encodes VDR proteins, which play an important role in inducing cell differentiation
and apoptosis as well as for the inhibition of cell proliferation and angiogenesis.
This suggests that through this mechanism apoptosis in cancer cells or precancerous
cells could be driven. This might prevent uncontrolled cell proliferation.
The breast epithelium appears to belong to the group of tissues affected by vitamin
D and VDR [6], [8], [9], [10]. Empirical research has already shown an inverse association between vitamin D serum
levels and the risk for breast cancer [10], [11], [12], [13], [14], [15], [16]. However, other studies found no correlation [17], [18], [19]. Results of studies which predominantly focused on vitamin D intake through nutrition
and its impact on breast cancer risk were also inconsistent [20].
Through its impact on breast epithelial cells it is possible that vitamin D might
not just have a direct impact on breast cancer risk but also on breast density. A
higher mammographic breast density is caused by an increased proportion of epidermal
and stromal cells compared to adipose tissue. As an increased proliferation of stromal
and epithelial tissue does not only result in higher breast density but is also associated
with a greater potential for malignant transformation [21], there is reason to suspect that there could be an association between higher breast
density and increased risk for breast cancer. Based on these considerations, some
research teams have already begun to focus on the consequences of daily vitamin D
supplements on mammographic breast density as a factor for breast cancer risk. Unfortunately,
the results were again very mixed [22], [23], [24], [25], [26]. As far as the authors of this paper know, there has been little research into the
effect of 25-hydroxyvitamin D serum levels on mammographic breast density – and the
results were again highly inconsistent [26], [27].
Therefore, the aim of this cross-sectional study of 984 subjects was to investigate
to what extent vitamin D serum levels correlated with mammographic breast density
and which additional factors could potentially influence both of these parameters.
Details of the study together with a comparison of women with malignant with women
with benign findings in their mammography have already been published [28].
Patients and Methods
Following a positive vote by the Ethics Committee of the Technical University of Munich,
1104 women aged from 20 to 88 years were recruited into the study. Of these, 984 underwent
mammography to determine breast density. The patient population consisted of asymptomatic
women, women with suspicious findings on palpation, and women with a genetic risk
of developing breast cancer over the longer term of at least 30 percent, including
carriers of BRCA1/BRCA2 gene mutations [29].
Women who were pregnant, women who were breastfeeding, women who had undergone breast
augmentation with implants, and women who had a history of breast cancer or who had
undergone breast surgery for B3 lesions were excluded from the study.
Patient characteristics and history
A short questionnaire-based interview was carried out to obtain the patientʼs general
medical and gynecological history and to record any known risk factors for breast
cancer along with the patientʼs lifestyle and eating habits. Collected data included
information on age and body mass index (BMI), alcohol and nicotine intake and any
chronic disease. The documented data also included information on reproductive factors
such as age at menarche, pregnancies, menopausal status, and hormone replacement therapy
and whether the patient had had an ovariectomy. Women were classified as post-menopausal
according to the WHO definition if their last spontaneous menstruation had occurred
at least twelve months previously and cessation of bleeding had been caused by the
loss of or decrease in ovarian endocrine function and was not the result of hysterectomy.
Previous breast surgeries and any familial history of breast or ovarian cancer were
recorded and particulars regarding the intake of vitamin D-rich food (dairy products,
fish, etc.), the intake of calcium and vitamin D preparations as well as physical
activity, the length of time spent outside, and the use of sun screen products were
also recorded.
After completing the interview 7.5 ml venous blood was taken to determine serum concentrations
of 25(OH)D, calcium, phosphate and creatinine. Vitamin D levels were determined using
VD3 (Vitamin D3) RIA kits (DiaSorin S. p. A.).
To avoid strong seasonal fluctuations which peak during the summer months the test
subjects were exclusively recruited from October to June.
Mammography
Two-plane mammography was carried out using standard mammography units. Mammography
was not done for the purposes of the study, and in most patients (n = 920) it was
usually performed on the day of the interview. The images were evaluated with regard
to architectural distortion, micro-calcifications and focal lesions and classified
using BI-RADS categories 0–6. Breast parenchymal density was classified into categories
1 to 4 using the ACR classification.
All mammograms were created using the same technology, stored on digital plates and
evaluated by two separate investigators. All participants in the study had standard
two-plane mammography (cranio-caudal [CC] and mediolateral oblique [MLO]), with additional
special imaging done as required. Two radiologists evaluated the mammographic density
using the BI-RADS Standard Atlas. This classifies breast density into four groups:
ACR 1 (almost entirely fatty breast tissue or glandular tissue < 25 %), ACR 2 (scattered
areas of fibroglandular density or 25–50 % glandular tissue), ACR 3 (heterogeneously
dense breast or 50–75 % glandular tissue) and ACR 4 (extremely dense breast or glandular
tissue > 75 %).
Abnormal mammographic findings were evaluated by histopathology and biopsy results
were provided in writing.
Statistical analysis
SPSS was used to calculate the measures of locality and variation of the individual
characteristics as well as their correlation with mammographic density based on the
ACR classification and vitamin D serum levels. T-tests for independent samples were
used to investigate statistically significant differences in mean values for 25(OH)D.
Multinomial logistic regression analysis was used to investigate the impact of different
factors on breast density. Parameters which already correlated with ACR in descriptive
statistical analysis and which also showed a significant correlation with ACR in regression
analysis were summarized as “main confounders” in Model 1 ([Table 3]). This included age, menopausal status and BMI as well as 25(OH)D levels to investigate
the central question of this study. All subsequent logistic models were based on Model
1 and took other additional potential confounders into consideration. These included
blood parameters, eating habits and lifestyle, prior breast surgery, familial risk,
reproductive parameters and chronic disease. The resulting regression coefficients
(B) correspond to logarithmic odds. Odds are calculated by dividing the probability
of occurrence (p) of an event by its converse probability (1-p). In our analysis,
1-p corresponded to the probability that the respective independent variable was in
ACR category 4. Probability of occurrence p described the probability of the variable
being in ACR categories 1, 2 or 3. The level of significance was set as α = 0.05.
Table 1 Mean vitamin D levels by the time of blood sampling by exogenous supply of vitamin
D, weight category, extent of sports activity, time spent outdoors, and indication
for biopsy based on mammographic findings. The mean 25(OH)D levels are shown together
with their standard deviations (SD) and the number of study participants (n) in the
individual subgroups.
|
Characteristics
|
Subgroups
|
25(OH)D (SD)
|
N
|
|
Patient population
|
|
17.2 (7.5)
|
984
|
|
Time of blood sampling
|
October – December
|
19.4 (7.7)
|
371
|
|
January – March
|
15.5 (6.9)
|
384
|
|
April – June
|
16.6 (7.1)
|
229
|
|
Intake of vitamin D preparations
|
no
|
16.5 (7.1)
|
838
|
|
yes
|
23.3 (7.7)
|
87
|
|
BMI
|
underweight
|
15.9 (7.9)
|
16
|
|
normal weight
|
18.0 (7.8)
|
602
|
|
pre-obesity
|
16.3 (6.4)
|
261
|
|
obese
|
14.8 (7.1)
|
105
|
|
Sports activity
|
rarely
|
15.2 (7.0)
|
271
|
|
sometimes
|
16.7 (7.4)
|
329
|
|
often
|
19.1 (7.5)
|
384
|
|
Time spent outdoors
|
rarely
|
14.7 (6.7)
|
347
|
|
sometimes
|
17.7 (7.4)
|
405
|
|
often
|
20.1 (7.5)
|
232
|
|
Histological investigation
|
no
|
17.4 (7.5)
|
882
|
|
yes
|
15.7 (6.8)
|
102
|
Table 2 Distribution of mammographic breast density (ACR classification) according to menopausal
status, years since menopause, BMI class for the pre- und post-menopausal groups and
overall; the quantiles of vitamin D levels in the pre- and post-menopausal groups
and overall; exogenous vitamin D intake; and according to the indication for biopsy
based on mammographic findings. The table shows the percentage of participants in
the respective breast density category (from ACR 1 [low density] to 4 [high density])
in the individual subgroups and the absolute sample size of the respective subgroup
(n).
|
Characteristics
|
Menopausal status
|
Subgroups
|
ACR 1
|
ACR 2
|
ACR 3
|
ACR 4
|
n
|
|
Patient population
|
pre-menopausal
|
|
2.9 %
|
33.0 %
|
46.4 %
|
17.7 %
|
412
|
|
post-menopausal
|
|
12.4 %
|
57.2 %
|
26.6 %
|
3.8 %
|
572
|
|
total
|
|
8.4 %
|
47.1 %
|
34.9 %
|
9.7 %
|
984
|
|
Years since menopause
|
≤ 5
|
9.9 %
|
49.1 %
|
31.7 %
|
9.3 %
|
161
|
|
6–10
|
10.7 %
|
57.9 %
|
28.9 %
|
2.5 %
|
121
|
|
11–15
|
14.5 %
|
62.7 %
|
20.9 %
|
1.8 %
|
110
|
|
16–20
|
14.5 %
|
56.6 %
|
27.7 %
|
1.2 %
|
83
|
|
> 20
|
14.4 %
|
63.9 %
|
20.6 %
|
1.0 %
|
97
|
|
BMI
|
pre-menopausal
|
underweight
|
|
10.0 %
|
40.0 %
|
50.0 %
|
10
|
|
normal weight
|
0.3 %
|
25.4 %
|
51.5 %
|
22.7 %
|
295
|
|
pre-obese
|
2.9 %
|
49.3 %
|
46.4 %
|
1.4 %
|
69
|
|
obese
|
23.7 %
|
68.4 %
|
7.9 %
|
|
38
|
|
post-menopausal
|
underweight
|
|
33.3 %
|
66.7 %
|
|
6
|
|
normal weight
|
6.2 %
|
52.8 %
|
34.9 %
|
6.2 %
|
307
|
|
pre-obese
|
16.7 %
|
62.5 %
|
19.3 %
|
1.6 %
|
192
|
|
obese
|
29.9 %
|
64.2 %
|
6.0 %
|
|
67
|
|
total
|
underweight
|
|
18.8 %
|
50.0 %
|
31.3 %
|
16
|
|
normal weight
|
3.3 %
|
39.4 %
|
43.0 %
|
14.3 %
|
602
|
|
pre-obese
|
13.0 %
|
59.0 %
|
26.4 %
|
1.5 %
|
261
|
|
obese
|
27.6 %
|
65.7 %
|
6.7 %
|
|
105
|
|
25(OH)D level
|
pre-menopausal
|
< 5
|
|
33.3 %
|
50.0 %
|
16.7 %
|
6
|
|
5–9
|
2.7 %
|
37.0 %
|
43.8 %
|
16.4 %
|
73
|
|
10–19
|
3.8 %
|
30.7 %
|
44.3 %
|
21.2 %
|
212
|
|
20–29
|
2.0 %
|
36.0 %
|
51.0 %
|
11.0 %
|
100
|
|
≥ 30
|
|
28.6 %
|
52.4 %
|
19.0 %
|
21
|
|
post-menopausal
|
< 5
|
|
66.7 %
|
33.3 %
|
|
6
|
|
5–9
|
16.4 %
|
52.7 %
|
29.1 %
|
1.8 %
|
55
|
|
10–19
|
12.7 %
|
59.7 %
|
23.0 %
|
4.7 %
|
300
|
|
20–29
|
11.4 %
|
55.4 %
|
30.3 %
|
2.9 %
|
175
|
|
≥ 30
|
11.1 %
|
50.0 %
|
33.3 %
|
5.6 %
|
36
|
|
total
|
< 5
|
|
50.0 %
|
41.7 %
|
8.3 %
|
12
|
|
5–9
|
8.6 %
|
43.8 %
|
37.5 %
|
10.2 %
|
128
|
|
10–19
|
9.0 %
|
47.7 %
|
31.8 %
|
11.5 %
|
512
|
|
20–29
|
8.0 %
|
48.4 %
|
37.8 %
|
5.8 %
|
275
|
|
≥ 30
|
7.0 %
|
42.1 %
|
40.4 %
|
10.5 %
|
57
|
|
Vitamin D preparations
|
pre-menopausal
|
no
|
2.8 %
|
32.2 %
|
46.2 %
|
18.8 %
|
388
|
|
yes
|
9.2 %
|
45.5 %
|
45.5 %
|
|
11
|
|
post-menopausal
|
no
|
13.1 %
|
58.4 %
|
23.8 %
|
4.7 %
|
450
|
|
yes
|
11.8 %
|
47.4 %
|
39.5 %
|
1.3 %
|
76
|
|
total
|
no
|
8.4 %
|
46.3 %
|
34.1 %
|
11.2 %
|
838
|
|
yes
|
11.5 %
|
47.1 %
|
40.2 %
|
1.1 %
|
87
|
|
Biopsy
|
pre-menopausal
|
no
|
3.3 %
|
34.8 %
|
43.8 %
|
18.1 %
|
365
|
|
yes
|
|
19.1 %
|
66.0 %
|
14.9 %
|
47
|
|
post-menopausal
|
no
|
12.8 %
|
56.9 %
|
26.1 %
|
4.3 %
|
517
|
|
yes
|
9.1 %
|
60.0 %
|
30.9 %
|
|
55
|
|
total
|
no
|
8.8 %
|
47.7 %
|
33.4 %
|
10.0 %
|
882
|
|
yes
|
4.9 %
|
41.2 %
|
47.1 %
|
6.9 %
|
102
|
Table 3 Regression coefficients for the effect of various influencing factors on mammographic
breast density using the ACR (American College of Radiologists) classification and
eight different logistic regression models. The table shows the regression coefficients
and standard error (in brackets). Model 1: adjusted for the main parameters: BMI (continuous), age (continuous), post-menopause
(reference: pre-menopause), 25(OH)D (continuous). Model 2: pre-menopausal patient group: adjusted for BMI (continuous), age (continuous), 25(OH)D
(continuous). Model 3: post-menopausal patient group: adjusted for BMI (continuous), age (continuous), 25(OH)D
(continuous). Model 4: breast parameters: listed independent variables: main parameter, BI-RADS 1/2 (reference:
BI-RADS 5/6); independent variables not listed here: BI-RADS 0, BI-RADS 3/4, mastitis,
breast surgery, benign breast changes, breast biopsy, familial risk for breast and
ovarian cancer. Model 5: gynecological parameters: independent variables listed: main parameter, bilateral
ovariectomy (reference: no ovariectomy); independent variables not listed here: age
at menarche, HRT, hysterectomy. Model 6: eating habits: independent variables listed: main parameter, intake of vitamin D
preparations (reference: no intake of vitamin D preparation); independent variables
not listed here: intake of fish/milk/yoghurt/cheese, intake of Ca preparations. Model 7: pre-menopausal cohort: confounders the same as in Model 6 with the exception of post-menopause.
Model 8: post-menopausal cohort: confounders the same as in Model 6 with the exception of
post-menopause.
|
Confounder
|
ACRa
|
Models
|
|
Model 1
|
Model 2
|
Model 3
|
Model 4
|
Model 5
|
Model 6
|
Model 7
|
Model 8
|
|
a: ACR 4 was used as the reference category.
***, **, *: statistically the regression coefficient does not equal zero at a significance
level of 0.1 or 1 or 5 %.
|
|
Constant term
|
1
|
− 21.037*** (1.773)
|
− 24.574*** (4.846)
|
− 18.645*** (3.053)
|
− 41.642*** (2.136)
|
− 20.434*** (1.927)
|
− 21.030*** (2.231)
|
− 28.870*** (7.300)
|
− 20.146*** (3.601)
|
|
2
|
− 13.064*** (1.357)
|
− 11.824*** (1.945)
|
− 12.521*** (2.736)
|
− 12.302*** (1.842)
|
− 12.650*** (1.491)
|
− 12.504*** (1.730)
|
− 12.422*** (2.368)
|
− 13.172*** (3.224)
|
|
3
|
− 6.518*** (1.274)
|
− 5.757*** (1.720)
|
− 7.912** (2.738)
|
− 6.301*** (1.741)
|
− 6.499*** (1.405)
|
− 6.291*** (1.655)
|
− 5.727** (2.106)
|
− 7.532* (3.226)
|
|
BMI
|
1
|
0.555***
(0.059)
|
0.753*** (0.101)
|
0.426*** (0.092)
|
0.596*** (0.062)
|
0.555*** (0.060)
|
0.557*** (0.068)
|
0.808*** (0.133)
|
0.459*** (0.101)
|
|
2
|
0.421*** (0.054)
|
0.505*** (0.072)
|
0.316*** (0.088)
|
0.431*** (0.055)
|
0.425*** (0.054)
|
0.421*** (0.061)
|
0.509*** (0.084)
|
0.330*** (0.096)
|
|
3
|
0.227*** (0.052)
|
0.271*** (0.067)
|
0.149 (0.089)
|
0.231*** (0.054)
|
0.228*** (0.053)
|
0.219*** (0.060)
|
0.266*** (0.079)
|
0.142 (0.098)
|
|
Age
|
1
|
0.092*** (0.025)
|
0.068 (0.074)
|
0.145*** (0.035)
|
0.110*** (0.027)
|
0.097*** (0.026)
|
0.090*** (0.028)
|
0.110 (0.099)
|
0.137*** (0.039)
|
|
2
|
0.063** (0.020)
|
0.007 (0.030)
|
0.120*** (0.032)
|
0.071*** (0.021)
|
0.066** (0.021)
|
0.060** (0.023)
|
− 0.006 (0.035)
|
0.114*** (0.036)
|
|
3
|
0.043* (0.019)
|
0.007 (0.027)
|
0.097** (0.032)
|
0.050* (0.021)
|
0.045* (0.021)
|
0.035 (0.022)
|
− 0.002 (0.032)
|
0.085* (0.036)
|
|
Post-menopause
|
1
|
1.203* (0.597)
|
|
|
1.020 (0.615)
|
1.272* (0.633)
|
0.992 (0.675)
|
|
|
|
2
|
0.745 (0.421)
|
|
|
0.676 (0.425)
|
0.570 (0.458)
|
0.378 (0.476)
|
|
|
|
3
|
0.068 (0.411)
|
|
|
0.013 (0.415)
|
0.003 (0.448)
|
− 0.029 (0.470)
|
|
|
|
25(OH)D
|
1
|
0.042 (0.024)
|
0.071 (0.053)
|
0.045 (0.037)
|
0.044 (0.025)
|
0.043 (0.024)
|
0.072* (0.029)
|
0.097 (0.076)
|
0.055 (0.044)
|
|
2
|
0.038* (0.018)
|
0.041 (0.022)
|
0.041 (0.033)
|
0.038* (0.018)
|
0.038* (0.018)
|
0.065** (0.022)
|
0.073** (0.028)
|
0.049 (0.040)
|
|
3
|
0.037* (0.017)
|
0.031 (0.020)
|
0.048 (0.033)
|
0.037* (0.017)
|
0.038* (0.017)
|
0.055** (0.021)
|
0.064* (0.026)
|
0.031 (0.040)
|
|
BI-RADS 1/2
|
1
|
|
|
|
20.486*** (0.731)
|
|
|
|
|
|
2
|
|
|
|
− 0.518 (1.032)
|
|
|
|
|
|
3
|
|
|
|
− 0.565 (0.972)
|
|
|
|
|
|
Bilateral ovariectomy
|
1
|
|
|
|
|
18.546*** (0.598)
|
|
|
|
|
2
|
|
|
|
|
18.524*** (0.368)
–
|
|
|
|
|
3
|
|
|
|
|
|
|
|
|
|
Intake of vitamin D preparations
|
1
|
|
|
|
|
|
1.533 (1.344)
|
20.265*** (3.539)
|
0.737 (1.543)
|
|
2
|
|
|
|
|
|
1.684 (1.221)
|
20.096*** (0.942)
|
0.974 (1.435)
|
|
3
|
|
|
|
|
|
2.407* (1.229)
|
–
|
1.892 (1.467)
|
Results
The mean serum concentration of 25-OH vitamin D for the total patient population was
17.2 ng/ml. Two of three women (n = 652) were vitamin D deficient (< 20 ng/ml) and
only 6 % had a physiologically adequate 25-hydroxyvitamin D supply of at least 30 ng/ml.
The lowest mean vitamin D levels were measured in March and the highest were measured
in November (14.65 ng/ml ± 6.61 SD vs. 20.78 ng/ml ± 9.17 SD). As expected, women
who took daily vitamin D preparations had higher vitamin D serum levels than women
who took no supplements (23.29 ng/ml ± 7.66 SD vs. 16.53 ng/ml ± 7.14 SD). There was
an inverse U-shaped association between 25(OH)D and BMI, whereby women of normal weight
had the highest vitamin D serum levels; the difference between patients of normal
weight and obese patients was highly significant (18.04 ng/ml ± 7.79 SD vs. 14.81 ng/ml
± 7.12 SD; t = 4.23, p < 0.001). The amount of sports and the time spent outdoors
showed an almost linear correlation with 25(OH)D. Women who underwent histological
work-up following their mammogram (n = 102) had average vitamin D levels which were
1.6 ng/ml lower than women who did not have a biopsy (15.67 ng/ml ± 6.81 SD vs. 17.37 ng/ml
± 7.51 SD; t = 2.37, p = 0.019). However, the histological findings did not correlate
with 25(OH)D levels (cf. [Table 1]).
Mammographic breast density (ACR classification)
The majority of the patient population had an intermediate breast density (ACR 2:
n = 463; ACR 3: n = 343). The percentage of women with high breast density declined
significantly following menopause; as the number of years since menopause increased,
the percentage of women with a lower breast density increased.
An inverse correlation was also found between ACR and BMI: while every third underweight
woman had a very high breast density (ACR 4), none of the obese women had a breast
density classified as ACR 4. High breast densities (ACR 3 or 4) were found significantly
more often in pre-menopausal women for whom biopsy was indicated compared to women
who did not require histological work-up (80.9 vs. 61.9 %). Participants with very
high breast density tended to have lower 25(OH)D levels than participants with an
average breast density (ACR 4: 15.91 ng/ml ± 7.84 SD; ACR 3: 17.74 ng/ml ± 7.77 SD;
t = 2.03, p = 0.044). But after further subdivision according to menopausal status,
this difference was no longer significant ([Fig. 1]). Regular intake of vitamin D supplements suggested a reduction in breast density
(cf. [Table 2]).
Fig. 1 Association between vitamin D levels and breast density in pre- and post-menopausal
women. Median vitamin D levels and 95 % confidence intervals for the respective ACR
categories differentiated according to the menopausal status of the 984 study participants
(n = 412 pre-menopausal, n = 572 post-menopausal). Of the pre-menopausal participants
2.9 % had ACR 1, 33.0 % had ACR 2, 46.4 % had ACR 3 and 17.7 % had ACR 4. Of the post-menopausal
women 12.4 % had ACR 1, 57.2 % had ACR 2, 26.6 % had ACR 3 and 3.8 % had ACR 4.
Overall, a total of 386 women – of whom 94 % were post-menopausal at the time of being
interviewed by questionnaire – reported having previously taken hormones as part of
hormone replacement therapy during menopause. At the time of participating in the
study only 111 women were taking hormones. Current hormone intake was not found to
be correlated with breast density in this patient subgroup, possibly because breast
density decreased significantly with time since menopause.
Both previous hormone therapy and ongoing hormone therapy at the time of the study
were associated with higher vitamin D levels compared to women who had never had hormone
therapy. The difference in serum levels for women receiving hormone therapy at the
time of the study was significantly lower compared to levels for women not currently
receiving hormone therapy (difference 0.71 ng/ml) and was associated with a significantly
bigger error bar than for women with previous hormone treatment compared to women
with no previous hormone intake (difference 1.64 ng/ml).
Multivariate logistic model
A multivariate logistic model with ACR as the dependent variable was used to calculate
the impact of the possible main confounders “BMI”, “age”, “menopausal status” and
“25(OH)D” on mammographic breast density (Model 1 in [Table 3]). With increasing age and BMI, low breast density was found significantly more often
than high breast density (ACR 1 vs. ACR 4, p < 0.001, respectively). A regression
coefficient B of 0.555 for BMI means that it is e0.555 = 1.742 times more probable
that a woman will have a very low breast density (ACR 1) rather than a high breast
density (ACR 4) when BMI is increased by one unit. Post-menopausal women had a greater
probability of having a very low breast density (ACR 1) compared to pre-menopausal
women (B = 1.203 for p = 0.044). Higher vitamin D levels were more likely in women
with medium breast density (ACR 2 and 3) compared to women with high breast density
(p = 0.032 and p = 0.028). The respective regression coefficients of 25(OH)D were
approximately the same for ACR groups 1 to 3 (B = 0.042; 0.038; 0.037). There were
only slight differences in vitamin D levels between these 3 categories, but all 3
categories had higher vitamin D levels compared to the high breast density group ([Table 3]).
After separately analyzing pre- and post-menopausal women, post-menopausal BMI and
age were found to be inversely correlated with breast density (p < 0.001). 25-Hydroxyvitamin
D did not appear to have an effect on post-menopausal breast density. This was in
contrast to the findings for pre-menopausal test subjects where BMI was negatively
associated with breast density (p < 0.001), but not age. Higher vitamin D levels were
not found significantly more often in women with medium breast density (ACR 2) compared
to women with ACR 4 (p = 0.060), possibly due to the smaller size of this group (n = 412).
After adjusting for all recorded breast changes and for familial risk for breast and
ovarian cancer according to Meindlʼs findings [29], BI-RADS 1 or 2 was unsurprisingly found to be associated significantly more often
with ACR 1 than BI-RADS categories 5 or 6 (p < 0.001).
Of all the other gynecological parameters studied, only a history of bilateral ovariectomy
was found to be associated with significantly lower breast density (ACR 1: B = 18.55,
ACR 2: B = 18.52 with p < 0.001, respectively).
Effect of exogenous vitamin D intake
When the eating habits of the test subjects were included in the analysis, the daily
intake of vitamin D preparations in the overall patient population was only weakly
associated with lower breast density (ACR 3: B = 2.41, p = 0.05). But if pre- and
post-menopausal women were analyzed separately, pre-menopausal women with regular
intake of vitamin D preparations had a significantly lower breast density compared
to pre-menopausal women who did not take vitamin D supplements (p < 0.001). In pre-menopausal
women, age was not correlated with breast density. In post-menopausal women, the intake
of vitamin D supplements appeared to have no effect on ACR in contrast to age which
appeared to strongly affect breast density. Neither vitamin D-containing nutrition,
blood parameters (calcium, phosphate, creatinine), sports, the use of sun screen and
the amount of time spent outside nor alcohol and smoking and the number of pregnancies
and age at pregnancy were found to be unambiguously associated with breast density.
When the figures were controlled for vitamin D-containing nutrition including the
intake of vitamin D supplements, the negative association between 25(OH)D and ACR
was heightened. The results of all other regression models on the association between
vitamin D and breast density were very similar to those of Model 1.
Discussion
25(OH)D levels of less than 10 ng/ml are considered deficient [30], [31]. Since 2010 many countries have raised the previous minimum value for normal vitamin
D levels from 20 to 30 ng/ml. It is not yet clear to what extent this could have an
impact on cancer prevention or other pleiotropic effects of vitamin D [30].
In our study only 6 % of participants had sufficient vitamin D supply (≥ 30 ng/ml),
and just under two thirds had low vitamin D levels (< 20 ng/ml). Although vitamin
D deficits were far more common in our patient population compared to the percentages
described in other European studies, those studies often differed strongly from our
study in their choice of patient population or study criteria [32], [33], [34], [35].
As cutaneous vitamin D synthesis depends very much on the incidence angle of the sunʼs
rays [36], the seasonal fluctuations of vitamin D levels found in our study population are
unsurprising, with the lowest levels recorded in March (mean: 14.65 ng/ml) and the
highest levels found in October and November (20.36 and 20.78 ng/ml). The quantitative
time spent outside was highly positively correlated with 25(OH)D levels, while the
intake of vitamin D-rich food was not at all correlated with 25(OH)D levels. In people
below the age of 60 years, cutaneous synthesis is the main supplier of vitamin D with
a share of up to 90 %; food plays a subordinate role for vitamin D balance [37], [38]. This appears to be different in countries in which nutrition such as dairy products
and cereals are fortified with vitamin D. A systematic review by OʼDonnell et al.
[39] reported a significantly positive impact on vitamin D levels of products with vitamin
D supplements. In our study population, regular intake of vitamin D preparations was
associated with a mean elevation of vitamin D serum levels by 40 %.
Frequent sports activities were also found to correlate positively with vitamin D
levels, as sports are often played outdoors [40], [41] and there is a negative correlation between sports and BMI [42], [43]. BMI is also being discussed as an influencing factor for 25(OH)D levels as both
severely underweight and obese women expose less skin to the sun and are more likely
to avoid outdoor activities [44], [45]. The results of our study also appear to indicate an inverse association between
BMI and vitamin D. Adipose tissue stores vitamin D and can therefore deplete the amount
of 25(OH)D present in blood, contributing to decreased serum levels of vitamin D [46]. Possible genetic connections are also being discussed. A recent study by Vimaleswaran
et al., which included 42 024 participants, showed that vitamin D deficiency occurred
more commonly in people with obesity-specific genetic variants [47].
Our analyses showed an inverse association between BMI and breast density. In addition
to breast density, BMI is also counted as one of the most important risk factors for
post-menopausal breast cancer [48]. Adipose tissue, which produces estrogens through aromatization, promotes cell proliferation
and mutations and increases both breast density and the risk of breast cancer [49], [50]. In women with higher BMI the total percentage of body fat as well as the percentage
of breast fat is elevated, which also explains the inverse association between BMI
and breast density [51], [52].
The post-menopausal reduction of mammographic density, which was also found in our
data, can be explained by the decreasing number of epithelial and stromal cells following
menopause [53], [54]. Unsurprisingly therefore, bilateral ovariectomy was found to be highly significantly
associated with lower breast density. The association between breast density and risk
for breast cancer could also be genetic [55], [56], [57].
Association between vitamin D and breast density
In our analysis, regular intake of vitamin D preparations was associated slightly
more (just reaching significance) with intermediate breast density (ACR 3) than with
high breast density (ACR 4) (p = 0.05; [Table 3]: Model 6). The intake of vitamin D-rich food did not appear to be relevant for breast
density. Other studies which examined a possible association between vitamin D and
breast density generally only looked at the intake of vitamin D from nutrition using
specific questionnaires but did not look at vitamin D levels in serum. In post-menopausal
women there was usually no correlation [23], [24], [58], [59], [60], while the breast density of pre-menopausal women appeared to be significantly inversely
associated with vitamin D intake [22], [58], [59], [61]. But as these studies used percentage breast tissue density instead of the ACR classification
used in our study, only limited comparisons are possible. As far as the authors of
this study could determine, only one other study has also used the ACR classification
to determine breast density [25]. That study reported a marginally significant inverse relation between ACR and vitamin
D intake in women with a high familial risk of breast and ovarian cancer. In our study
population, pre-menopausal women who took regular vitamin D supplements were more
likely to have low breast densities and this likelihood was highly significant, but
with a relatively large regression coefficient (> 20) compared to pre-menopausal women
with no vitamin D intake, although the sample size of this evaluation was considerably
smaller than that used in Model 6. No such association was found for post-menopausal
women.
Whether the intake of vitamin D preparations could lead to a reduction in breast density
at least in pre-menopausal women is therefore still not clear.
Our findings indicate a conditional inverse connection between 25(OH)D and breast
density. Although following multivariate adjustment the presence of low vitamin levels
made the probability of high breast density (ACR 4) significantly more likely, vitamin
D serum levels for the respective ACR groups 1–3 barely differed from one another.
It could therefore be assumed that vitamin D levels are only important in women with
high breast density. The few studies on the effect of vitamin D on mammographic density
mainly examined post-menopausal participants and found almost no associations [26], [62], [63], [64]. The recently published analysis by Bertrand et al. of 835 pre-menopausal women
showed significantly higher percent breast tissue densities for women with vitamin
D levels in the highest 25(OH)D quartile compared to levels in the lowest 25(OH)D
quartile [27]. After including the respective risk for breast cancer, they found that higher vitamin
D serum levels in women with high breast density were correlated with a lower risk
for breast cancer. No such association was found for women with low to medium breast
density. This is in accordance with our findings.
When separate regression analyses were done for pre- and post-menopausal women, the
results differed strongly from those of Model 1; while BMI and ACR correlated inversely
in both pre-menopausal and post-menopausal women and this inverse correlation was
highly significant, age was inversely associated with breast density only in post-menopausal
women. In pre-menopausal participants, higher vitamin D serum levels tended to be
associated with lower breast density (p = 0.060 for ACR 2 vs. ACR 4); no such association
was found in post-menopausal women. The results therefore indicate that the effect
of vitamin D serum levels is primarily pre-menopausal. After menopause, 25(OH)D appears
to play a subordinate role for breast density. Because breast density significantly
decreases as the time since menopause increases, no significant effect of current
hormone therapy on breast density was found in the small study population receiving
HT (n = 111).
Conclusion
Although a strong antiproliferative and immunomodulatory effect is ascribed to vitamin
D, up to now the findings of studies on possible connections between vitamin D and
breast cancer have been very heterogeneous. The possible association with breast density
as an established risk factor for breast cancer is still controversially discussed.
The results of our large cross-sectional study with just under 1000 participants indicate
an inverse relationship between vitamin D and mammographic density which appears to
be strongly dependent on menopausal status. Even after multivariate adjustment for
different factors influencing breast density, the incidence of low breast density
was significantly higher for all pre-menopausal women who had either high 25(OH)D
levels or who took regular vitamin D supplements ([Table 3], Models 2 and 7). After menopause, vitamin D was not correlated with ACR; BMI and
age however were significantly correlated inversely with breast density.
More studies will be necessary to confirm the hypothesis of a primarily pre-menopausal
relationship between vitamin D and breast density. In addition to possible long-term
effects of 25(OH)D and vitamin D supplements on breast density, a follow-up study
could record the incidence of breast cancer. The results of such studies could possibly
bring us one step closer to successfully preventing breast cancer.
Irrespective of how important vitamin D is for the prevention of breast cancer, there
is now a broad consensus that vitamin D insufficiency is very common in the general
population and should not be underestimated. The well-known and significant negative
consequences of this on bone density and the possible connection to a number of other
diseases make it all the more important to investigate and understand the precise
mechanism of action of vitamin D and its derivatives in the human body in further
studies.
