Background:

Some studies have shown that cadmium (Cd) is associated with breast cancer risk. One hypothesis is that Cd has estrogen-like properties. This case-control study investigated the association between breast cancer risk and blood Cd (BCd) levels.

Methods:

All breast cancers in the Malmö Diet and Cancer cohort were identified through linkage to the Swedish Cancer Registry, baseline (1991–1996) through 2014. Two controls per case were selected from the same cohort. BCd was analyzed at baseline. Associations were analyzed using logistic regression.

Results:

Mean BCd was 0.51 μg/L among 1,274 cases and 0.46 among 2,572 controls. There was an overall increased risk of breast cancer [OR, 1.18; 95% confidence interval (CI), 1.05–1.36] per μg/L of BCd. An increased risk was, however, only found at high BCd [OR, 1.34 (95% CI, 1.05–1.73)] for BCd more than 1.20 μg/L. The group with the highest BCd was mainly smokers. A spline indicated that at BCd less than 1.0 μg/L, the OR was not increased. The association with BCd was stronger in current smokers and at body mass index (BMI) above 25, while no modification due to receptor status was found.

Conclusions:

The results indicated increased risk of breast cancer only for high Cd exposure, which occurred mainly among smokers. This made it difficult to disentangle the effects of smoking and Cd, despite inclusion of smoking habits in the models.

Impact:

This study provides support for reducing Cd exposure through smoking cessation and dietary choice. On the population level, preventive measures against Cd pollution are warranted.

It has been suggested that the heavy metal cadmium (Cd) is associated with increased risk of breast cancer, but results from epidemiologic studies are inconsistent, as reported below.

Many established risk factors for breast cancer are hormone- and pregnancy-related (1). It has been suggested that Cd is a metalloestrogen — a metal that binds to estrogen receptors (2), and mimics the actions of estrogen. The Cd concentration in breast cancer tissue seems to be higher than in normal breast tissue (3). Cd is proestrogenic in vivo (2, 4, 5). Cd is classified as a group 1 carcinogen by the International Agency for Research on Cancer (IARC) due to increased risk of lung cancer.

Nonsmokers in Sweden are exposed to Cd mainly through food (e.g., cereals, potatoes, vegetables, ref. 6). Cd in tobacco smoke contributes to higher blood Cd (BCd) levels (7). The uptake of Cd is higher among persons with iron deficiency, probably due to an upregulation of a common receptor in the intestine (8).

For dietary Cd (assessed using questionnaires), some studies found an association with breast cancer. In the population-based Swedish Mammography Cohort, the RR was 1.21, [95% confidence interval (CI), 1.07–1.36], for highest versus lowest tertile of dietary Cd intake (9). An Italian cohort study also showed an increased risk [HR, 1.54 (95% CI, 1.06–2.22)] for highest versus lowest quintile of dietary Cd (10). However, there are also studies showing no clear association (11–14).

Cd has a long half-life in the body (10–40 years) and exposure can be assessed using biomarkers of Cd in urine (UCd) or BCd (7). Using UCd, two case–control studies reported a doubled risk for breast cancer [OR, 2.29; 95% CI, 1.3–4.2 for UCd >0.58 vs. <0.26 μg/g creatinine (15), and OR, 2.69; 95% CI, 1.07–6.78, for UCd >0.60 vs. < 0.22 μg/g creatinine (16)]. There are, however, also studies showing no association (17, 18). Fewer studies have used BCd, but a recent study was negative, reporting a relative risk of breast cancer of 0.84 per μg/L (95% CI, 0.69–1.01; ref. 19). In summary, the literature on Cd and breast cancer is inconsistent.

The aim of this study is to investigate and quantify the association between breast-cancer risk and Cd exposure (measured by BCd in prediagnostic samples), using cases and controls from the large Swedish Malmo Diet and Cancer cohort with a follow-up time of around 20 years.

Study population

The population-based Malmö Diet and Cancer study (MDCS) has been described elsewhere (20, 21), but in short all men and women living in Malmö, Sweden, and born during 1923 to 1950 were invited between 1991 and 1996 to participate. A total of 17,035 women were recruited. Part of the cohort was reexamined during 2007 to 2012 (22).

Outcome

Through the Swedish personal identification number, the women were linked to the Swedish Cancer Registry, and followed until first breast cancer diagnosis, death, emigration, or end of follow-up (December 31, 2014). Emigration was identified through the National Statistics registers. Breast cancer events, including carcinomia in situ, were assigned as International Classification of Diseases 7 (ICD7) = 170 (the Swedish cancer register uses the seventh revision of the ICD as the main code).

After excluding prevalent breast cancer cases (diagnosis before baseline), there were 1,292 women with a diagnosis registered after baseline and up to December 31, 2014. From the same cohort, 2,584 female controls were selected, frequency-matched for time of inclusion in MDCS (± 6 months), date of birth (± 6 months), and selected among those still alive at diagnosis of cases. This selection approach is called incidence density sampling (23).

Invasive tumors diagnosed between 1991 and 2004 were pathologically reevaluated regarding estrogen receptor (ER) and progesterone receptor (PgR) status by one senior pathologist (24, 25). For tumors from 2005 and onwards, this information was available from pathologic reports. ER and PgR were considered positive at a cutoff of >10% positively stained nuclei.

Baseline covariates

During baseline visits, anthropometric measurements were performed, and lifestyle factors, socioeconomic status, and medication were assessed using established questionnaires. Age at menarche, age at menopause, number of children were assessed. Menopausal status at baseline was defined as postmenopausal if menstruation had stopped more than 2 years prior to baseline, or if the woman had undergone bilateral oophorectomy. The status was defined as perimenopausal if the menstruation had stopped less than 2 years prior to baseline. For 2% of the women, the above information was not available and then the woman was classified as perimenopausal if she was 42 to 55 years at baseline, and as postmenopausal if she was over 55 at baseline (26). Smoking status was assessed at baseline (never smokers, former smokers, or current smokers). Number of smoking years and number of cigarettes per day were used to calculate pack-years. Diet data were collected through a 7-day menu book and a dietary questionnaire (27, 28), and fiber intake (g per day) was used in the present study. Leisure-time physical activity was classified from 18 questions (29, 30), and a summary score combined the intensity factors for each activity with the time spent on the activity, and the score was categorized into quartiles. Highest education was categorized into ≤9 years of schooling, 10 to 12 years, or university. Socioeconomic index was based on type of work (manual, nonmanual, employer, other). Information on oophorectomy was obtained from the Swedish National in-patient register. Fasting blood samples were drawn and stored in a biobank.

Exposure assessment

The fasting blood samples at baseline were drawn in heparin tubes and erythrocytes were stored frozen in cryovials. To measure the Cd concentrations, thawed erythrocytes were used, for all cases and controls in this study. In short, Cd was analyzed by inductively coupled plasma mass spectrometry with an octopole reaction system, operated in the helium collision cell mode (details published previously in ref. 31). The Cd level in blood was calculated by multiplying with hematocrit. Samples were analyzed in different rounds. Four lots of external quality control (QC) samples were included (Seronorm Trace Elements Whole Blood, Sero AS). The results compared with recommended limits were 0.34 ± 0.02 μg/L (N = 70) versus 0.32–0.40 μg/L (L-1, Lot no.1103128), 5.7 ± 0.18 μg/L (N = 70) versus 5.4–6.2 μg/L (L-2, Lot no.1103129), 0.27 ± 0.014 μg/L (N = 35) versus 0.17–0.40 μg/L (L-1, Lot no.1406263), 5.19 ± 0.181 μg/L (N = 35) versus 4.00–6.02 μg/L (L-2, Lot no.1406264). The imprecision was 9.6% (the coefficient of variation for 50 duplicate samples) and the detection limit, calculated as three times the SD of the blank, was 0.02 μg/L. None of the samples were below the limit of detection. An interlaboratory comparison showed good agreement (Pearson correlation coefficient 0.99, slope 1.04).

Statistical analysis

Multiple logistic regression was used (both conditional and unconditional) and BCd was included both as a continuous and categorical variable. Nonlinear associations were explored by specifying a spline (B-spline basis) in the EFFECT statement of the PROC LOGISTIC procedure (SAS Institute Inc). Comparisons between cases and controls were made using t tests or χ2 tests.

Some of the risk factors for breast cancer, considered to have strong scientific support using the GRADE system (32), are late menopause, hormone replacement therapy, mutations in the BRCA1 and BRCA2 genes, and obesity (for postmenopausal breast cancer; Supplementary Table S1). Lifestyle factors are alcohol consumption (positive association), long breast-feeding, and physical activity (negative association; ref. 33).

Three regression models were fitted. In model 1, similar factors as in a recent study on breast cancer and the micronutrient selenium (26) were included, namely body mass index (BMI <20, 20–25, 25–30, or >30 kg/m2), age at first childbirth (<20, 20–25, 25–30, or >30 years), age at menarche (years), age at baseline (years), education (3 categories as described above), socioeconomic index (4 categories as described above), marriage (yes/no), number of children (0, 1, 2, 3, 4, or more), oophorectomy (yes/no), menopausal status (pre, peri, post), hormone replacement therapy (HRT, yes/no), use of oral contraceptives (yes/no), alcohol consumption (No alcohol, <15 g/day, 15–30 g/day, >30 g/day, infrequent use), smoking habits (current, former, never). Model 2 was based on the directed acyclic graph (DAG) in Supplementary Material, Supplementary Fig. S1, and adjusted for BMI, socioeconomic index, physical activity (4 categories as described above), menopausal status, age at first childbirth, age at baseline, and HRT. This was our main model. The third model included risk factors for breast cancer and determinants for Cd exposure, such as smoking and fiber consumption (metal concentrations are generally higher in whole grains than in refined products; ref. 34). This Model 3 included number of children, age at first childbirth, BMI, alcohol consumption, HRT, physical activity, and late menopause (above 50) as well as smoking and fiber consumption (g/MJ, proxy for Cd intake through food) and socioeconomic index.

First, the exposure-outcome curve was estimated (i) under assumption of a constant risk increase (per μg/L) and (ii) allowing the association to be nonlinear, through both categorization and a spline function. The associations in different receptor status groups were compared. Secondly, subgroup analyses were performed, based on smoking status, BMI, age at event, and for postmenopausal women. Between-group heterogeneity was assessed using the Cochran Q statistic, and tested by a χ2 test.

The study was approved by the Ethics Committee at Lund University, Sweden (# 2015/283 and 2017/859).

BCd could be determined in 3,846 women (2,572 controls and 1,274 cases, of which 99 were in situ). The cases and controls had the same mean age at baseline, 56 years, and similar distributions on many variables, but HRT and nonmanual work was more common among the cases (Table 1). In the Supplementary Material, the available breast cancer risk factors (from Supplementary Table S1) are presented by BCd level among controls (Supplementary Table S2).

Table 1.

Baseline information on 1,274 incident breast cancer cases and 2,572 matched controls from the MDCS (baseline 1991–1996, followed until December 31, 2014).

n
CasesControls(cases;controls)aPb
 Mean (min–max) Mean (min–max)   
Blood cadmium (μg/L)c 0.51 (0.05–5.0) 0.46 (0.04–4.3) 1,274;2,572 0.02 
Birth year (1923–1950) (1923–1950) 1,274;2,572 0.96 
Age at baseline 56.2 (44.7–73.5) 56.2 (44–73) 1,274;2,572 0.98 
Age at menarche 13.5 (10.0–20.0) 13.6 (9–22) 1,189;2,518 0.15 
Age at first child 24.8 (12.0–45.0) 24.7 (11–45) 1,009;2,174 0.34 
Age at menopause 49.3 (31.0–64.0) 48.8 (28–61) 711;1,540 0.03 
Age at diagnosis 66.8 (45–91) 1,274; - 
Height (cm) 164.3 (139–180) 163.9 (139–184) 1,200;2,536 0.06 
Weight (kg) 69.0 (40.0–124.0) 68.1 (39–135) 1,200;2,536 0.02 
BMI, kg/m2 25.6 (16.2–46.1) 25.4 (15.2–50.2) 1,200;2,536 0.13 
Hemoglobin (g/L) 137 (102–175) 136 (78–182) 1,273;2,572 0.44 
Energy (kcal/day) 2,079 (628–4,025) 2,065 (637–5,744) 1,203;2,544 0.43 
Fiber (g/day) 19.4 (2.6–55.9) 19.6 (5.5–64.7) 1,203;2,544 0.51 
 % (n) % (n)   
Parity   1,176;2,485 0.02 
 0 14% (166) 13% (311)   
 1 19% (223) 22% (538)   
 2 47% (547) 42% (1,049)   
 3 15% (182) 17% (432)   
 ≥4 5% (58) 6% (155)   
Alcohol consumption   1,200;2,570 0.003 
 No alcohol 5% (63) 6% (153)   
 <15 g/day 64% (763) 66% (1,678)   
 15–30 g/day 15% (185) 14% (364)   
 >30 g/day 4% (50) 2% (50)   
 Infrequent use 12% (139) 12% (293)   
Hormone replacement therapy   1,197;2,530 <0.0001 
 Yes 26% (315) 18% (465)   
 No 74% (882) 82% (2,065)   
Married/cohabiting   1,200;2,539 0.41 
 Yes 68% (813) 69% (1,754)   
 No 32% (387) 31% (785)   
Socioeconomic indexd   1,187;2,519 <0.0001 
 Manual 33% (396) 40% (1,001)   
 Non-manual 61% (720) 52% (1,321)   
 Employer 6% (71) 8% (197)   
Menstrual statuse   1,200;2,540 0.63 
 Pre 31% (370) 29% (746)   
 Peri 8% (100) 9% (224)   
 Post 61% (730) 62% (1,570)   
Education   1,197;2,530 0.02 
 ≤9 years 66% (796) 69% (1,750)   
 10–12 years 7% (78) 8% (197)   
 University 27% (323) 23% (583)   
Smoking   1,200;2,539 0.48 
 Never 42% (508) 44% (1,128)   
 Current 28% (338) 27% (695)   
 Former 30% (354) 28% (716)   
Use oral contraceptives   1,199;2,540 0.19 
 Ever 54% (649) 52% (1,317)   
 Never 46% (550) 48% (1,223)   
Oophorectomy, bilateral   1,200;2,540 0.85 
 Yes 1% (17) 1% (38)   
 No 99% (1,183) 99% (2,502)   
Physical activityf    0.13 
 Low 25% (304) 25% (629) 1,209;2,525  
 Moderate 52% (627) 49% (1,241)   
 High 23% (278) 26% (655)   
 % (n) % (n)   
Receptor statusg   (only cases)  
 ER+ 89% (762)   
 ER 11% (90)   
 PgR+ 60% (502)   
 PgR 40% (323)   
n
CasesControls(cases;controls)aPb
 Mean (min–max) Mean (min–max)   
Blood cadmium (μg/L)c 0.51 (0.05–5.0) 0.46 (0.04–4.3) 1,274;2,572 0.02 
Birth year (1923–1950) (1923–1950) 1,274;2,572 0.96 
Age at baseline 56.2 (44.7–73.5) 56.2 (44–73) 1,274;2,572 0.98 
Age at menarche 13.5 (10.0–20.0) 13.6 (9–22) 1,189;2,518 0.15 
Age at first child 24.8 (12.0–45.0) 24.7 (11–45) 1,009;2,174 0.34 
Age at menopause 49.3 (31.0–64.0) 48.8 (28–61) 711;1,540 0.03 
Age at diagnosis 66.8 (45–91) 1,274; - 
Height (cm) 164.3 (139–180) 163.9 (139–184) 1,200;2,536 0.06 
Weight (kg) 69.0 (40.0–124.0) 68.1 (39–135) 1,200;2,536 0.02 
BMI, kg/m2 25.6 (16.2–46.1) 25.4 (15.2–50.2) 1,200;2,536 0.13 
Hemoglobin (g/L) 137 (102–175) 136 (78–182) 1,273;2,572 0.44 
Energy (kcal/day) 2,079 (628–4,025) 2,065 (637–5,744) 1,203;2,544 0.43 
Fiber (g/day) 19.4 (2.6–55.9) 19.6 (5.5–64.7) 1,203;2,544 0.51 
 % (n) % (n)   
Parity   1,176;2,485 0.02 
 0 14% (166) 13% (311)   
 1 19% (223) 22% (538)   
 2 47% (547) 42% (1,049)   
 3 15% (182) 17% (432)   
 ≥4 5% (58) 6% (155)   
Alcohol consumption   1,200;2,570 0.003 
 No alcohol 5% (63) 6% (153)   
 <15 g/day 64% (763) 66% (1,678)   
 15–30 g/day 15% (185) 14% (364)   
 >30 g/day 4% (50) 2% (50)   
 Infrequent use 12% (139) 12% (293)   
Hormone replacement therapy   1,197;2,530 <0.0001 
 Yes 26% (315) 18% (465)   
 No 74% (882) 82% (2,065)   
Married/cohabiting   1,200;2,539 0.41 
 Yes 68% (813) 69% (1,754)   
 No 32% (387) 31% (785)   
Socioeconomic indexd   1,187;2,519 <0.0001 
 Manual 33% (396) 40% (1,001)   
 Non-manual 61% (720) 52% (1,321)   
 Employer 6% (71) 8% (197)   
Menstrual statuse   1,200;2,540 0.63 
 Pre 31% (370) 29% (746)   
 Peri 8% (100) 9% (224)   
 Post 61% (730) 62% (1,570)   
Education   1,197;2,530 0.02 
 ≤9 years 66% (796) 69% (1,750)   
 10–12 years 7% (78) 8% (197)   
 University 27% (323) 23% (583)   
Smoking   1,200;2,539 0.48 
 Never 42% (508) 44% (1,128)   
 Current 28% (338) 27% (695)   
 Former 30% (354) 28% (716)   
Use oral contraceptives   1,199;2,540 0.19 
 Ever 54% (649) 52% (1,317)   
 Never 46% (550) 48% (1,223)   
Oophorectomy, bilateral   1,200;2,540 0.85 
 Yes 1% (17) 1% (38)   
 No 99% (1,183) 99% (2,502)   
Physical activityf    0.13 
 Low 25% (304) 25% (629) 1,209;2,525  
 Moderate 52% (627) 49% (1,241)   
 High 23% (278) 26% (655)   
 % (n) % (n)   
Receptor statusg   (only cases)  
 ER+ 89% (762)   
 ER 11% (90)   
 PgR+ 60% (502)   
 PgR 40% (323)   

Note: Cases and controls were matched on year of birth and time of inclusion into study.

aMissing (cases and controls): Parity 98 and 87, Alcohol 74 and 34, Hormone replacement therapy 77 and 42, Married/cohabiting 74 and 33, Socioeconomic index 87 and 53, Menstrual status 74 and 32, Education 77 and 42, Smoking 74 and 33, Oral contraceptives 75 and 32, Oophorectomy 74 and 32, Physical activity 65 and 47.

bComparing cases and controls using a t test (for the continuous variables) or a χ2 test (for the categorical variables).

cThe median was 0.28 μg/L among both cases and controls.

dThere were no women who had socioeconomic index “Other”.

eMenstrual status, constructed by JM: Post if menstruation had stopped more than 2 years prior to baseline, or if bilateral oophorectomy; Peri if menstruation had stopped less than 2 years prior to baseline. In 2% of the women, neither information was available; therefore, if they were ≥42 years at baseline, they were classified as perimenopausal (42–55 years) or postmenopausal (>55 years). Women were classified as premenopausal status if not post or peri.

fLeisure-time physical activity during the preceding year was classified based on 18 questions, Li et al. 2009, Taylor et al. 1978. A summary score for all activities was obtained by combining intensity factors for each activity with the time spent on the activity, and the score was categorized, based on quartiles, into 4 groups (Q1, Q2, Q3, Q4). The group with the lowest score, Q1, is classified as Low, Q2-Q3 is classified as Moderate, and Q4 as High.

gCancer in situ and bilateral cancer have been excluded. 422 cases did not have classification of estrogen receptor status, and 449 did not have classification of progesterone receptor status.

Association with Cd in the entire group

There were 3,559 women with complete data on all the variables (1,147 cases and 2,412 controls). In the conditional logistic regression, only the cases with at least one control are included in the analysis, leaving 3,290 observations. BCd was included as a linear effect, or categorized on the basis of the BCd cutoffs 0.40, 0.80, and 1.20 μg/L.

When all observations were used (n = 3,559), the OR was approximately 1.2 per μg/L of BCd (Table 2). But excluding the 5% highest BCd observations resulted in a lower risk estimate [OR, 1.14; 95% CI, 0.91–1.44 per μg/L (using Model 2 Supplementary Table S3)]. Among those with the highest BCd values, >1.20 μg/L, the OR was 1.34 (95% CI, 1.05–1.73; Table 3).

Table 2.

The ORs, for breast cancer, presented per 1 μg/L of BCd, and per 0.36 μg/L (interquartile range) of BCd.

Conditional logistic regressiona(Unconditional) logistic regressionb
OR (95% CI)pOR per IQROR (95% CI)pOR per IQR
Model 1 1.18 (0.97–1.42) 0.09 1.06 1.20 (1.00–1.44) 0.05 1.07 
Model 2 1.18 (1.02–1.36) 0.02 1.06 1.20 (1.04–1.38) 0.01 1.06 
Model 3 1.17 (0.97–1.42) 0.10 1.06 1.20 (1.00–1.44) 0.05 1.07 
Conditional logistic regressiona(Unconditional) logistic regressionb
OR (95% CI)pOR per IQROR (95% CI)pOR per IQR
Model 1 1.18 (0.97–1.42) 0.09 1.06 1.20 (1.00–1.44) 0.05 1.07 
Model 2 1.18 (1.02–1.36) 0.02 1.06 1.20 (1.04–1.38) 0.01 1.06 
Model 3 1.17 (0.97–1.42) 0.10 1.06 1.20 (1.00–1.44) 0.05 1.07 

Note: Model 1: BCd, parity, age, BMI, HRT, education, age at first child, alcohol, smoking, age at menarche, socioeconomic index, married, oophorectomy, menopausal status, oral contraceptives. Model 2: BCd, BMI, physical activity, menopausal status, socioeconomic index, age, age at first child, HRT. Model 3: BCd, parity, age at first child, late menopause, HRT, BMI, alcohol, physical activity, smoking, fiber consumption, socioeconomic index. P values for two-sided tests are presented.

Abbreviation: IQR, interquartile range.

aConditional logistic regression based on 1,006+136 strata (3,290 observations, 1,142 events).

b(Unconditional) logistic regression based on 3,559 observations, 1,147 events.

Table 3.

The ORs, for breast cancer, presented for BCd categorized by cutoffs 0.40, 0.80, and 1.20 μg/L, with the 95% CI in parentheses.

BCd (μg/L) categories
0.041–0.400.40–0.800.80–1.201.20–4.997
Mean; median 0.22; 0.21 0.56; 0.54 0.98; 0.97 1.81; 1.61 
Cases 751 196 84 116 
Conditional logistic regression (1,006+136 strata, 3,290 observations, 1,142 events) 
Model 1 1.04 (0.82–1.31) 1.08 (0.74–1.57) 1.31 (0.92–1.86) 
P 0.76 0.68 0.13 
Model 2 1.04 (0.85–1.27) 1.10 (0.82–1.47) 1.30 (1.003–1.69) 
P 0.71 0.54 0.05 
Model 3 1.04 (0.83–1.32) 1.09 (0.75–1.59) 1.31 (0.92–1.86) 
P 0.72 0.64 0.13 
(Unconditional) logistic regression (3,559 observations, 1,147 events) 
Model 1 1.02 (0.81–1.27) 1.05 (0.74–1.50) 1.32 (0.94–1.85) 
P  0.88 0.79 0.10 
Model 2 1.04 (0.86–1.26) 1.08 (0.82–1.43) 1.34 (1.05–1.73) 
P  0.70 0.59 0.02 
Model 3 1.04 (0.83–1.30) 1.06 (0.74–1.50) 1.33 (0.95–1.86) 
P  0.72 0.77 0.10 
BCd (μg/L) categories
0.041–0.400.40–0.800.80–1.201.20–4.997
Mean; median 0.22; 0.21 0.56; 0.54 0.98; 0.97 1.81; 1.61 
Cases 751 196 84 116 
Conditional logistic regression (1,006+136 strata, 3,290 observations, 1,142 events) 
Model 1 1.04 (0.82–1.31) 1.08 (0.74–1.57) 1.31 (0.92–1.86) 
P 0.76 0.68 0.13 
Model 2 1.04 (0.85–1.27) 1.10 (0.82–1.47) 1.30 (1.003–1.69) 
P 0.71 0.54 0.05 
Model 3 1.04 (0.83–1.32) 1.09 (0.75–1.59) 1.31 (0.92–1.86) 
P 0.72 0.64 0.13 
(Unconditional) logistic regression (3,559 observations, 1,147 events) 
Model 1 1.02 (0.81–1.27) 1.05 (0.74–1.50) 1.32 (0.94–1.85) 
P  0.88 0.79 0.10 
Model 2 1.04 (0.86–1.26) 1.08 (0.82–1.43) 1.34 (1.05–1.73) 
P  0.70 0.59 0.02 
Model 3 1.04 (0.83–1.30) 1.06 (0.74–1.50) 1.33 (0.95–1.86) 
P  0.72 0.77 0.10 

Note: Model 1: BCd, parity, age, BMI, HRT, education, age at first child, alcohol, smoking, age at menarche, socioeconomic index, married, oophorectomy, menopausal status, oral contraceptives. Model 2: BCd, BMI, physical activity, menopausal status, socioeconomic index, age, age at first child, HRT. Model 3: BCd, parity, age at first child, late menopause, HRT, BMI, alcohol, physical activity, smoking, fiber consumption, socioeconomic index. There were 3,559 observations and 1,147 events. P values for two-sided tests are presented.

The association was further explored by a spline. The left panel of Fig. 1 shows an increasing OR above 1 μg/L, and the right panel shows that the OR is significantly increased above BCd 1.2 μg/L.

Figure 1.

The association between breast cancer and BCd, using Model 2 and (unconditional) logistic regression. The spline (A) presents the OR with 95% confidence limits, in relation to BCd = 0.10 μg/L. The categories (B; cutoffs 0.4, 0.8, and 1.2) presents the ORs compared with the lowest category (BCd<0.40 μg/L). Note that the scales are different.

Figure 1.

The association between breast cancer and BCd, using Model 2 and (unconditional) logistic regression. The spline (A) presents the OR with 95% confidence limits, in relation to BCd = 0.10 μg/L. The categories (B; cutoffs 0.4, 0.8, and 1.2) presents the ORs compared with the lowest category (BCd<0.40 μg/L). Note that the scales are different.

Close modal

As the uptake of Cd is higher among persons with iron deficiency, hemoglobin was used as a marker for iron status (neither ferritin nor serum-Fe was available). Persons with hemoglobin <120 g/L were considered to be anemic. Excluding 95 women with anemia did not change the OR, compared with Tables 2 and 3 (Supplementary Table S4A and S4B).

The ER status among invasive tumors (excluding cancer in situ and bilateral tumors) was classified as ER+ in 762 cases (Table 1). The median BCd was 0.28 μg/L in ER+ cases and 0.25 μg/L in the 90 ER cases. Sixty percent were classified as PgR+ (Table 1). The median BCd was 0.28 μg/L both for PgR+ and PgR cases. For the 762 ER-positive tumors (and their controls), the OR was 1.34 (95% CI, 0.98–1.84) for BCd > 1.20 μg/L, which is similar to the overall OR. The regression estimates for the small ER-negative group were unstable. For PgR-positive and -negative tumors, the CIs were wide: for BCd > 1.20 versus BCd < 0.40 μg/L; OR, 1.42 (95% CI, 0.96–2.10) for PgR+, and OR, 1.12 (95% CI, 0.69–1.82) for PgR (details in Supplementary Tables S5A and S5B).

Including only the invasive tumors and their controls gave OR = 1.33 (95% CI, 1.01–1.75) for BCd > 1.20 versus BCd < 0.40 μg/L (Supplementary Tables S6A and S6B), which was similar to the results based on all observations.

Subgroup analyses based on smoking status, BMI, menopausal status, and age at event

As mentioned above, smoking is an important determinant for BCd levels, and separate analyses were conducted for never-smokers, ex-smokers, and current smokers. The distribution of BCd is very different between the smoking groups, making it difficult to find a common categorization. When the cut-offs 0.40, 0.80, and 1.20 were used (resulting in 443, 266, and 42 events respectively in the reference category for each smoking group), only the smokers had high BCd values and an OR of 1.34 (95% CI, 0.85–2.12). Common BCd categories for the 3 smoking groups means that OR for BCd > 1.20 μg/L could not be estimated among never- and ex-smokers (Table 4, Supplementary Table S7A). When assuming a constant risk increase per μg/L of BCd, no association was seen among never-smokers [OR, 1.00 per μg/L of BCd (95% CI, 0.49–2.03)]. Among ex-smokers, OR = 1.12 per μg/L of BCd (95% CI, 0.54–2.34). Among current smokers the OR was 1.25 per μg/L of BCd (95% CI, 1.02–1.53; (Supplementary Table S7B). At the reexamination from 2007 to 2012, smoking history was reassessed. In a sensitivity analysis, only those with a consistent smoking status (those where status did not change) were included and cases diagnosed before the reexamination were excluded (163 events) together with their controls, leaving 2,281 individuals with a complete set of covariates. The results (Supplementary Table S8A and S8B) were slightly stronger than for the whole group, especially when comparing BCd>1.20 with BCd<0.40 μg/L [OR, 1.53 (95% CI, 1.09–2.14)]. Adjusting for pack-years attenuated the association with breast cancer slightly (see Fig. 2).

Table 4.

The ORs, for breast cancer, presented for BCd categorized by cutoffs 0.40, 0.80, and 1.20 μg/L, stratified by smoking.

BCd (μg/L)BCd meanCasesOR95% CIp
Never smokers 0.041–0.40 0.20 443   
n = 1,555 0.40–0.80 0.51 39 1.06 0.71–1.58 0.78 
485 cases 0.80–1.20 — —   
 1.20–5.00 — —   
Ex-smokers 0.041–0.40 0.23 266   
n = 1,013 0.40–0.80 0.52 62 0.91 0.64–1.28 0.58 
337 cases 0.80–1.20 0.94   
 1.20–5.00 —   
Smokers 0.041–0.40 0.26 42   
n = 991 0.40–0.80 0.60 95 1.04 0.66–1.65 0.86 
325 cases 0.80–1.20 0.98 77 1.04 0.65–1.68 0.86 
 1.20–5.00 1.82 111 1.34 0.85–2.12 0.20 
BCd (μg/L)BCd meanCasesOR95% CIp
Never smokers 0.041–0.40 0.20 443   
n = 1,555 0.40–0.80 0.51 39 1.06 0.71–1.58 0.78 
485 cases 0.80–1.20 — —   
 1.20–5.00 — —   
Ex-smokers 0.041–0.40 0.23 266   
n = 1,013 0.40–0.80 0.52 62 0.91 0.64–1.28 0.58 
337 cases 0.80–1.20 0.94   
 1.20–5.00 —   
Smokers 0.041–0.40 0.26 42   
n = 991 0.40–0.80 0.60 95 1.04 0.66–1.65 0.86 
325 cases 0.80–1.20 0.98 77 1.04 0.65–1.68 0.86 
 1.20–5.00 1.82 111 1.34 0.85–2.12 0.20 

Note: Estimated using (unconditional) logistic regression, model 2 (BCd, BMI, physical activity, menopausal status, socioeconomic index, age, age at first child, hormone replacement therapy). P values (two-sided tests) are presented.

Figure 2.

The association between breast cancer and BCd, using Model 2 and (unconditional) logistic regression. A, ORs, for BCd>1.20 vs. BCd<0.40 μg/L, where “LCL” and “UCL” denote lower and upper confidence limit, respectively, “Cases_1.20” is the number of events in the highest BCd category and mBCd denotes the mean BCd in the group. B, OR per μg/L of BCd, where “Cases” denotes the total number of events. In both graphs, “(a)” denotes analysis based only on smokers with information on pack-years. Cochran Q statistic was 6.04 [A, P = 0.42, based on 7 groups; “All” and “Current smoker (a)” excluded], and 3.85 [B, P = 0.87, based on 9 groups; “All” and “Current smoker (a)” excluded].

Figure 2.

The association between breast cancer and BCd, using Model 2 and (unconditional) logistic regression. A, ORs, for BCd>1.20 vs. BCd<0.40 μg/L, where “LCL” and “UCL” denote lower and upper confidence limit, respectively, “Cases_1.20” is the number of events in the highest BCd category and mBCd denotes the mean BCd in the group. B, OR per μg/L of BCd, where “Cases” denotes the total number of events. In both graphs, “(a)” denotes analysis based only on smokers with information on pack-years. Cochran Q statistic was 6.04 [A, P = 0.42, based on 7 groups; “All” and “Current smoker (a)” excluded], and 3.85 [B, P = 0.87, based on 9 groups; “All” and “Current smoker (a)” excluded].

Close modal

Among never-smokers, 53% had a BMI above 25 (16% had BMI>30), 37% among current smokers (10% had BMI>30), and 49% among ex-smokers (14% had BMI>30). The association between BCd and breast cancer was stronger among women with BMI above 25, see Fig. 2.

At baseline, 61% of the women were classified as having postmenopausal status and 8% as perimenopausal (Table 1). The median BCd levels among the postmenopausal women were very similar for the 730 cases (0.26 μg/L) and the 1,570 controls (0.27 μg/L). Compared with the whole group (n = 3,559), the OR was lower when analyzing the postmenopausal group (Fig. 2; complete results in Supplementary Table S9A and S9B).

The median age at event was 66 years (range 45.7–91.4). In a comparison based on the age at event of the cases (≤66 vs. >66 years at event), the OR was higher in the older group [OR, 1.80 (95% CI, 1.22–2.65); see Fig. 2].

In this Swedish population-based case–control study, we found an association between breast cancer risk and BCd, but only among women with BCd>1.20 μg/L. This association was mainly due to an association in smokers. At BCd<1.0 μg/L, the OR was not increased. Among the cases, only 8% were classified as perimenopausal at baseline, and more than 50% as postmenopausal, and in postmenopausal women adipose tissue is a source of estrogen production. Among women with BMI above 25 in our study, the OR was 1.50 (95% CI, 1.04–2.19) when comparing BCd>1.20 and BCd<0.40 μg/L. Only 11% of the invasive tumors were classified as ER and thus no clear conclusion could be made regarding differential risk due to estrogen receptor status, only that the ER+ group had higher BCd.

Cd is considered to have estrogen-like properties, (35), and to mimic estradiol by binding to the ER alpha and activating it (36–39). According to IARC, it is not yet known whether or not Cd promotes tumor growth by an estrogen-mediated mechanism (40). Other suggested mechanisms behind Cd as a carcinogen are disturbances of DNA repair and changes in DNA methylation patterns (40–42). Cd levels among women are often higher than for men, as a result of low iron stores (43). Previous epidemiologic studies, where the exposure assessment was based on dietary intake or UCd, have shown mixed results for Cd as a risk factor for breast cancer. Several case–control studies have reported increased risk: for highest versus lowest tertile of Cd intake, (44) reported OR = 1.23 (95% CI, 0.76–2.00), and for high versus low UCd levels, (45) reported OR = 6.05 (95% CI, 2.90–12.62) and (46) reported OR = 1.62 (95% CI, 1.19–2.21). U.S. studies compared UCd approximately above 0.6 to approximately below 0.24 μg/g creatinine and reported a pooled OR of 2.4 (95% CI, 1.55–3.72; ref. 47). But there are also studies that report no clear association with UCd (18, 48, 49), nor with dietary Cd (49).

Studies where the exposure was measured by BCd have also shown contradictory results. A Chinese case–control study reported higher BCd concentration among cases compared with controls (2.28 and 1.77 μg/L, respectively; ref. 50). A National Health and Nutrition Survey (NHANES)-based study indicated a higher self-reported prevalence of breast cancer, OR = 1.29 (95% CI, 0.73–2.28), among women with BCd>0.6 compared with BCd<0.3 μg/L (51). Our results, however, indicated a higher risk of breast cancer among women with BCd levels (above 1.2 μg/L), OR = 1.34 (95% CI, 1.05–1.73), which was not found in the meta-analysis by Gaudet, where the RR was 0.59 (95% CI, 0.39–0.91) when comparing BCd levels above 2 μg/L and below 0.49 μg/L (19). In our study, the BCd levels (median 0.28 μg/L) are considerably lower than in studies from China, where median blood concentrations around 2–3 μg/L was reported among patients with breast cancer, and median 1.77 μg/L among controls (50, 52).

Assessing Cd exposure using dietary questionnaires provide only a rough estimate of long-term exposure to Cd, (53), whereas both UCd and BCd are well established, and provide reliable measures of exposure and body burden (7, 54). The half-life of UCd is longer than that of BCd (7).

In our study, the median BCd was higher among ER-positive tumors (0.28 μg/L) than in ER-negative tumors (0.25 μg/L). Since most of the classified tumors were ER-positive (762 of 852), we had no power to compare the receptor groups. For the ER+ group, OR was 1.34 (95% CI, 0.98–1.84) for BCd>1.20 μg/L, which is similar to the results for all the tumors. Two previous case-control studies both reported stronger associations among ER-positive breast cancers: OR = 1.79 (1.30–2.46) when comparing UCd>0.33 with UCd<0.18 μg/g creatinine, (55), and using Cd intake, OR = 1.94 (95% CI, 1.04–3.63) when comparing high (31.5 μg/day) and low (21.4 μg/day) (44). For BCd>1.20 compared with BCd<0.40 μg/L, our study consistently showed higher (nonsignificant) risk estimates in the PgR+ group, compared with PgR. This was also reported in ref. 55, where OR was 1.88 (95% CI, 1.30–2.74), for UCd > 0.33 versus UCd < 0.18 μg/g creatinine. Also in a prospective study, the risk estimates (HR) were higher in the receptor-positive groups, although the heterogeneity was not significant (10).

A previous cohort study reported overrepresentation of ER-negative tumors among smokers and former smokers (56). In our study, however, the proportion of ER-negative tumors was only 9% among current and former smokers, and 12% among never smokers. If smokers are more likely to have ER-negative tumors, then the mechanism would not be through binding to the ER. Smoking is complicated, since it is an important source of Cd as mentioned above, and therefore a vast majority of the women with the highest BCd levels (>1.20 μg/L) were smokers, 311 of 322. For another estrogen-based cancer, endometrial cancer, cigarette smoking was associated with a reduced risk, especially among postmenopausal women (57). However, as obesity is a possible factor for endometrial cancer, (58), the reduced risk among smokers might be due to less adipose tissue and therefore decreased estrogen levels.

Smokers and nonsmokers differ regarding dietary habits, but the association between diet and breast cancer is not clear. For fat intake and breast cancer risk, a case–control study (nested within MDCS) found different associations depending on how fat intake was assessed (59, 60). According to a review (33), there is only limited evidence that intake of, for example, nonstarchy vegetables or dairy products, is related to breast cancer risk. It is difficult to assess how different diets in smokers and nonsmokers will affect associations between BCd and breast cancer risk. Smoking is not considered a risk factor according to recent reviews (33, 61), which would indicate that the increased BCd-related risk we found among smokers could be due to their high BCd levels.

The three covariate models mostly gave similar results, which indicates robustness. As smoking is a large source of Cd, it could be argued that models 1 and 3 imply an over-adjustment by including smoking. However, if we consider those results that had a P value for OR smaller than 0.10, the ORs are often slightly higher for models 1 and 3. Model 2 often has the narrowest CI, which is probably an effect of fewer variables included. Our study is a population-based case-control study, where the two matched controls were selected from the same cohort as the cases. Both unconditional and conditional logistic regression was used, where the latter uses the discordant pairs (the pairs where case and control have different exposure).

Some limitations should be acknowledged. The heparin tubes used for blood collection were not tested for possible contamination of Cd. The BCd concentration in this study are, however, consistent with results from other studies in Sweden, and no outlier with unexpectedly high concentrations were found. BCd and covariates were assessed at baseline, and may have changed before the breast cancer diagnosis date. Although BCd is affected partly by recent Cd exposure, it is also a good biomarker of body burden if smoking and dietary habits have not changed recently. For example, in US former smokers, the correlation with previous exposure (pack-years of smoking) was in fact higher for BCd than for UCd (62). Moreover, in this study, a sensitivity analysis based on only women with consistent smoking pattern, also showed an increased risk for BCd>1.20 μg/L [OR, 1.53 (95% CI, 1.09–2.14)]. The distribution of BCd is skewed with few high values, making the estimation of the exposure-risk curve difficult. In addition to including Cd as a continuous variable, we categorized Cd, and used a spline to illustrate the exposure–response relationship.

A strength is that we used prediagnostic exposure to avoid reverse causation. Bias was avoided by selecting all cases in the MDCS cohort, by selecting controls from the same cohort as the cases and by not having selection criteria related to the exposure (63). A further strength was that the controls were matched on birthday within +-6 months, ensuring similar age ranges. The cases were identified from the Swedish National Cancer Registry, which has shown a high level of completeness and a low underreporting for breast cancer (64).

In conclusion, the results provide support for a nonlinear dose–response relation between Cd exposure and breast cancer, with a threshold for increased risk at relatively high Cd exposure, which in the present cohort was the case mainly in smokers. On the individual level, Cd exposure can be reduced through smoking cessation. A reduction of Cd levels in common foods (cereals, potatoes) is warranted, for example, by using fertilizers with low Cd concentration.

E.M. Andersson reports grants from Swedish Research Council for Health, Working Life and Welfare (FORTE) during the conduct of the study. L. Barregard reports grants from The Sahlgrenska University Hospital, Region Västra Götaland under the ALF agreement during the conduct of the study. No disclosures were reported by the other authors.

E.M. Andersson: Software, formal analysis, funding acquisition, visualization, methodology, writing–original draft, project administration, writing–review and editing. M. Sandsveden: Investigation, writing–review and editing. N. Forsgard: Resources, investigation, writing–review and editing. G. Sallsten: Conceptualization, methodology, writing–review and editing. J. Manjer: Data curation, validation, methodology, writing–review and editing. G. Engström: Data curation, methodology, writing–review and editing. L. Barregard: Conceptualization, funding acquisition, methodology, writing–review and editing.

This study was funded by the Swedish Research Council for Health, Working Life and Welfare (FORTE, # 2017-00519), by the foundation “Allmänna sjukhusets i Malmö stiftelse för bekämpande av cancer”, and by The Sahlgrenska Univertsity Hospital, Region Västra Götaland under the ALF agreement (ALFGBG-924961).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Veronesi
U
,
Boyle
P
,
Goldhirsch
A
,
Orecchia
R
,
Viale
G
. 
Breast cancer
.
Lancet
2005
;
365
:
1727
41
.
2.
Byrne
C
,
Divekar
S
,
Storchan
G
,
Parodi
D
,
Martin
M
. 
Cadmium—a metallohormone?
Toxicol Appl Pharmacol
2009
;
238
:
266
71
.
3.
Strumylaite
L
,
Bogusevicius
A
,
Abdrachmanovas
O
,
Baranauskiene
D
,
Kregzdyte
R
,
Pranys
D
, et al
Cadmium concentration in biological media of breast cancer patients
.
Breast Cancer Res Treat
2011
;
125
:
511
7
.
4.
Ali
I
,
Damdimopoulou
P
,
Stenius
U
,
Adamsson
A
,
Mäkelä
S
,
Åkesson
A
, et al
Cadmium-induced effects on cellular signaling pathways in the liver of transgenic estrogen reporter mice
.
Toxicol Sci
2012
;
127
:
66
75
.
5.
Johnson
M
,
Kenney
N
,
Stoica
A
,
Hilakivi-Clarke
L
,
Singh
B
,
Chepko
G
, et al
Cadmium mimics the in vivo effects of estrogen in the uterus and mammary gland
.
Nat Med
2003
;
9
:
1081
4
.
6.
Swedish National Food Agency
.
Swedish Market Basket Survey 2015 – per capita-based analysis of nutrients and toxic compounds in market baskets and assessment of benefit or risk in Livsmedelsverkets rapportserie
.
Swedish National Food Agency (Livsmedelsverket)
; 
2017
.
7.
Nordberg
G
,
Nogawa
K
,
Nordberg
M
. 
Cadmium.
In:
Nordberg
G
,
Fowler
G
,
Nordberg
M
,
editors
.
Handbook on the toxicology of metals
.
Amsterdam, the Netherlands
:
Elsevier
; 
2015
.
p.
667
716
.
8.
Berglund
M
,
Akesson
A
,
Nermell
B
,
Vahter
M
. 
Intestinal absorption of dietary cadmium in women depends on body iron stores and fiber intake.
Environ Health Perspect
1994
;
102
:
1058
66
.
9.
Julin
B
,
Wolk
A
,
Bergkvist
L
,
Bottai
M
,
Åkesson
A
. 
Dietary cadmium exposure and risk of postmenopausal breast cancer: a population-based prospective cohort study
.
Cancer Res
2012
;
72
:
1459
66
.
10.
Grioni
S
,
Agnoli
C
,
Krogh
V
,
Pala
V
,
Rinaldi
S
,
Vinceti
M
, et al
Dietary cadmium and risk of breast cancer subtypes defined by hormone receptor status: a prospective cohort study
.
Epidemiology
2019
;
144
:
2153
60
.
11.
Adams
S
,
Newcomb
P
,
White
E
. 
Dietary cadmium and risk of invasive postmenopausal breast cancer in the VITAL cohort
.
Cancer Causes Control
2012
;
23
:
845
54
.
12.
Sawada
N
,
Iwasaki
M
,
Inoue
M
,
Takachi
R
,
Sasazuki
S
,
Yamaji
T
, et al
Long-term dietary cadmium intake and cancer incidence
.
Epidemiology
2012
;
23
:
368
76
.
13.
Adams
S
,
Quraishi
S
,
Shafer
M
,
Passarelli
M
,
Freney
E
,
Chlebowski
R
, et al
Dietary cadmium exposure and risk of breast, endometrial, and ovarian cancer in the Women's Health Initiative.
Environ Health Perspect
2014
;
122
:
594
600
.
14.
Eriksen
K
,
Halkjær
J
,
Sørensen
M
,
Meliker
J
,
McElroy
J
,
Tjønneland
A
, et al
Dietary cadmium intake and risk of breast, endometrial and ovarian cancer in Danish postmenopausal women: a prospective cohort study
.
PLoS One
2014
;
9
:
e100815
.
15.
McElroy
J
,
Shafer
M
,
Trentham-Dietz
A
,
Hampton
J
,
Newcomb
P
. 
Cadmium exposure and breast cancer risk
.
J Natl Cancer Inst
2006
;
98
:
869
73
.
16.
Gallagher
C
,
Chen
J
,
Kovach
J
. 
Environmental cadmium and breast cancer risk
.
Aging
2010
;
2
:
804
14
.
17.
Adams
S
,
Shafer
M
,
Bonner
M
,
LaCroix
A
,
Manson
J
,
Meliker
J
, et al
Urinary cadmium and risk of invasive breast cancer in the Women's Health Initiative
.
Am J Epidemiol
2016
;
183
:
815
23
.
18.
Eriksen
K
,
McElroy
J
,
Harrington
J
,
Levine
K
,
Pedersen
C
,
Sørensen
M
, et al
Urinary cadmium and breast cancer: a prospective danish cohort study
.
J Natl Cancer Inst
2017
;
109
:
7
.
19.
Gaudet
M
,
Deubler
E
,
Kelly
R
,
Diver
W
,
Teras
L
,
JM
H
, et al
Blood levels of cadmium and lead in relation to breast cancer risk in three prospective cohorts
.
Int J Cancer
2019
;
144
:
1010
6
.
20.
Berglund
G
,
Elmståhl
S
,
Janzon
L
. 
The Malmö Diet and Cancer Study. Design and feasibility
.
J Intern Med
1993
;
223
:
45
51
.
21.
Manjer
J
,
Carlsson
S
,
Elmstahl
S
,
Gullberg
B
,
Janzon
L
,
Lindstrom
M
, et al
The Malmo Diet and Cancer Study: representativity, cancer incidence and mortality in participants and non-participants
.
Eur J Cancer Prev
2001a
;
10
:
489
99
.
22.
Rosvall
M
,
Persson
M
,
Östling
G
,
Nilsson
P
,
Melander
O
,
Hedblad
B
, et al
Risk factors for the progression of carotid intima-media thickness over a 16-year follow-up period: the Malmö Diet and Cancer Study
.
Atherosclerosis
2015
;
239
:
615
22
.
23.
Richardson
D
. 
An incidence density sampling program for nested casecontrol analyses
.
Occup Environ Med
2004
;
61
:
e59
.
24.
Borgquist
S
,
Anagnostaki
L
,
Jirström
K
,
Landberg
G
,
Manjer
J
. 
Breast tumours following combined hormone replacement therapy express favourable prognostic factors
.
Int J Cancer
2007
;
120
:
2202
7
.
25.
Butt
S
.
Reproductive factors and breast cancer - parity, breastfeeding and genetic predisposition in relation to risk and prognosis
.
Lund, Sweden
:
Lund University
; 
2011
.
26.
Sandsveden
M
,
Manjer
J
. 
Selenium and breast cancer risk: a prospective nested case–control study on serum selenium levels, smoking habits and overweight
.
Int J Cancer
2017
;
141
:
1741
50
.
27.
Mattisson
I
,
Wirfalt
E
,
Gullberg
B
,
Berglund
G
. 
Fat intake is more strongly associated with life style factors than with socio-economic characteristics, regardless of energy adjustment approach
.
Eur J Clin Nutr
2001
;
55
:
452
61
.
28.
Wirfält
E
,
Mattisson
I
,
Johansson
U
,
Gullberg
B
,
Wallstrom
P
,
Berglund
G
. 
A methodological report from the Malmö Diet and Cancer study: development and evaluation of altered routines in dietary data processing
.
Nutr J
2002a
;
1
:
16
.
29.
Li
C
,
Aronsson
C
,
Hedblad
B
,
Gullberg
B
,
Wirfält
E
,
Berglund
G
. 
Ability of physical activity measurements to assess health-related risks
.
Eur J Clin Nutr
2009
;
63
:
1448
51
.
30.
Taylor
H
,
Jacobs
D
 Jr
,
Schucker
B
,
Knudsen
J
,
Leon
A
,
Debacker
G
. 
A questionnaire for the assessment of leisure time physical activities
.
J Chronic Dis
1978
;
31
:
741
55
.
31.
Fagerberg
B
,
Barregard
L
,
Sallsten
G
,
Forsgard
N
,
Ostling
G
,
Persson
M
, et al
Cadmium exposure and atherosclerotic carotid plaques–results from the Malmö diet and Cancer study
.
Environ Res
2015
;
136
:
67
74
.
32.
Kavanagh
B
. 
The GRADE system for rating clinical guidelines
.
PloS Med
2009
;
6
:
e1000094
.
33.
Continuous Update Project Expert Report, Diet, nutrition, physical activity and breast cancer, in CUP (Continous Update Project)
. 
2017
,
World Cancer Research Fund/American Institute for Cancer Research
.
34.
Thielecke
F
,
Nugent
A
. 
Contaminants in Grain-A major risk for whole grain safety?
Nutrients
2018
;
10
:
1213
.
35.
Aquino
N
,
Sevigny
M
,
Sabangan
J
,
Louie
M
. 
Role of cadmium and nickel in estrogen receptor signaling and breast cancer: metalloestrogens or not?
J Environ Sci Health C Environ Carcinog Ecotoxicol Rev
2012
;
30
:
189
224
.
36.
Byrne
C
,
Divekar
S
,
Storchan
G
,
Parodi
D
,
Martin
M
. 
Metals and breast cancer
.
J Mammary Gland Biol Neoplasia
2013
;
18
:
63
73
.
37.
Martin
M
,
Reiter
R
,
Pham
T
,
Avellanet
Y
,
Camara
J
,
Lahm
M
, et al
Estrogen-like activity of metals in MCF-7 breast cancer cells
.
Endocrinol
2003
;
144
:
2425
36
.
38.
Darbre
P
. 
Metalloestrogens: an emerging class of inorganic xenoestrogens with potential to add to the oestrogenic burden of the human breast
.
J Appl Toxicol
2006
;
26
:
191
7
.
39.
Stoica
A
,
Katzenellenbogen
B
,
Martin
M
. 
Activation of estrogen receptor-α by the heavy metal cadmium
.
Mol Endocrinol
2000
;
14
:
545
53
.
40.
International Agency for Research on Cancer
. 
Cadmium and cadmium compounds
.
In
:
Arsenic, Metals, Fibres, and Dusts
.
IARC Monographs on the Evaluation of Carcinogenic Risks to Humans Volume 100C
.
Lyon, France
:
IARC
; 
2012
.
41.
McElroy
J
,
Hunter
M
. 
Cadmium: a new risk factor for endometrial cancer?
Expert Rev Anticancer Ther
2019
;
19
:
355
8
.
42.
Morales
M
,
Derbes
R
,
Ade
C
,
Ortego
J
,
Stark
J
,
Deininger
P
, et al
Heavy metal exposure influences double strand break DNA repair outcomes
.
PLoS One
, 
2016
;
11
:
e0151367
.
43.
Vahter
M
,
Berglund
M
,
Akesson
A
,
Liden
C
. 
Metals and women's health
.
Environ Res
2002
;
88
:
145
55
.
44.
Itoh
H
,
Iwasaki
M
,
Sawada
N
,
Takachi
R
,
Kasuga
Y
,
Yokoyama
S
, et al
Dietary cadmium intake and breast cancer risk in Japanese women: a case–control study
.
Int J Hyg Environ Health
2014
;
217
:
70
77
.
45.
Nagata
C
,
Nagao
Y
,
Nakamura
K
,
Wada
K
,
Tamai
Y
,
Tsuji
M
, et al
Cadmium exposure and the risk of breast cancer in Japanese women
.
Breast Cancer Res Treat
2013
;
138
:
235
9
.
46.
Strumylaite
L
,
Kregzdyte
R
,
Bogusevicius
A
,
Poskiene
L
,
Baranauskiene
D
,
Pranys
D
. 
Association between cadmium and breast cancer risk according to estrogen receptor and human epidermal growth factor receptor 2: epidemiological evidence
.
Breast Cancer Res Treat
2014
;
145
:
225
32
.
47.
Lin
J
,
Zhang
F
,
Lei
Y
., 
Dietary intake and urinary level of cadmium and breast cancer risk: a meta-analysis
.
Cancer Epidemiol
2016
;
42
:
101
7
.
48.
García-Esquinas
E
,
Pollan
M
,
Tellez-Plaza
M
,
Francesconi
K
,
Goessler
W
,
Guallar
E
, et al
Cadmium exposure and cancer mortality in a prospective cohort: the strong heart study
.
Environ Health Perspect
2014
;
122
:
363
.
49.
Filippini
T
,
Torres
D
,
Lopes
C
,
Carvalho
C
,
Moreira
P
,
Naska
A
, et al
Cadmium exposure and risk of breast cancer: a dose-response meta-analysis of cohort studies
.
Environ Int
2020
;
142
:
105879
.
50.
Peng
L
,
Huang
Y
,
Zhang
J
,
Peng
Y
,
Lin
X
,
Wu
K
, et al
Cadmium exposure and the risk of breast cancer in Chaoshan population of southeast China
.
Environ Sci Pollut Res
2015
;
22
:
19870
8
.
51.
Wei
Y
,
Zhu
J
. 
Blood levels of endocrine-disrupting metals and prevalent breast cancer among US women
.
Med Oncol
2020
;
37
:
1
.
52.
He
Y
,
Peng
L
,
Huang
Y
,
Liu
C
,
Zheng
S
,
Wu
K
. 
Blood cadmium levels associated with short distant metastasis-free survival time in invasive breast cancer
.
Environ Sci Pollut Res
2017
;
24
:
28055
64
.
53.
Vacchi-Suzzi
C
,
Eriksen
K
,
Levine
K
,
McElroy
J
,
Tjønneland
A
,
Raaschou-Nielsen
O
, et al
Dietary intake estimates and urinary cadmium levels in Danish postmenopausal women
.
PLoS One
2015
;
10
:
e0138784
.
54.
Åkesson
A
,
Barregard
L
,
Bergdahl
I
,
Nordberg
G
,
Nordberg
M
,
Skerfving
S
. 
Non-renal effects and the risk assessment of environmental cadmium exposure.
Environ Health Perspect
2014
;
122
:
431
8
.
55.
Strumylaite
L
,
Kregzdyte
R
,
Bogusevicius
A
,
Poskiene
L
,
Baranauskiene
D
,
Pranys
D
. 
Cadmium exposure and risk of breast cancer by histological and tumor receptor subtype in white Caucasian women: a hospital-based case-control study
.
Int J Mol Sci
2019
;
20
:
3029
.
56.
Manjer
J
.
Smoking and breast cancer
.
Department of Community Medicine, Malmö University Hospital
:
Lund University, Sweden
; 
2001
;
125
.
57.
Zhou
B
,
Yang
L
,
Sun
Q
. 
Cigarette smoking and the risk of endometrial cancer:a meta-analysis
.
Am J Med
2008
;
121
:
501
8
.
58.
Kalliala
I
,
Markozannes
G
,
Gunter
M
. 
Obesity and gynaecological and obstetric conditions: umbrella review of the literature
.
BMJ
2017
;
359
:
j4511
.
59.
Wirfält
E
,
Mattisson
I
,
Gullberg
B
,
Johansson
U
,
Olsson
H
,
Berglund
G
. 
Postmenopausal breast cancer is associated with high intakes of x6 fatty acids (Sweden)
.
Cancer Causes Control
2002b
;
13
:
883
93
.
60.
Wirfält
E
,
Vessby
B
,
Mattisson
I
,
Gullberg
B
,
Olsson
H
,
Berglund
G
. 
No relations between breast cancer risk and fatty acids of erythrocyte membranes in postmenopausal women of the Malmö Diet Cancer cohort (Sweden)
.
Eur J Clin Nutr
2004
;
58
:
761
77
.
61.
Cancercentrum
,
[Swedish] Nationellt vårdprogram Bröstcancer
.
Stockholm, Sweden
:
Regionala cancercentrum i samverkan
; 
2019
.
62.
Hecht
E
,
Arheart
K
,
Lee
D
,
Hennekens
C
,
Hlaing
W
. 
Interrelation of cadmium, smoking, and cardiovascular disease (from the National Health and Nutrition Examination Survey)
.
Am J Cardiol
2016
;
118
:
204
9
.
63.
Schulz
K
,
Grimes
D
. 
Case-control studies: research in reverse
.
Lancet
2002
;
359
:
431
4
.
64.
Barlow
L
,
Westergran
K
,
Holmberg
L
,
Talbäck
M
. 
The completeness of the Swedish Cancer Registery sample survey for year 1998
.
Acta Oncol
2008
;
48
:
27
33
.