Serum calcium, tumor size, and hormone receptor status in women with untreated breast cancer

Abstract

Elevated serum levels of calcium are frequently observed in advanced breast cancer, but data on serum calcium and breast cancer characteristics at the time of breast cancer diagnosis are limited. We conducted a cross-sectional study of 555 women with newly-diagnosed, untreated breast cancer in North Dakota. We examined the relationship between tumor size, serum calcium and other clinical characteristics of breast tumors, including age and hormone receptor status, using multiple linear regressions. Tumors that were estrogen receptor negative tended to be associated with higher serum calcium levels (p = 0.07). We observed a significant positive correlation between tumor volume and serum calcium levels (adjusted for patient age, body mass index, hormonal receptors, stage at diagnosis, and grade). The association between tumor volume and serum calcium was limited to post-menopausal women. Our finding that postmenopausal women with larger breast tumors had significantly higher serum calcium levels is consistent with a calciotropic effect of early breast cancer in postmenopausal women.

Keywords: :

• serum calcium
• albumin
• hormone receptors
• breast cancer
• epidemiology

Introduction

Elevated levels of calcium in serum are frequently observed in advanced breast cancer.¹ The hypercalcemia observed in metastatic breast cancer is commonly the result of osteolysis caused by bony metastases. However, elevated serum calcium levels may occur in the absence of bony metastases and are caused by the production by the breast tumors of humoral factors (humoral hypercalcemia of malignancy), such as parathyroid hormone related protein (PTHrP).² PTHrP, the principal agent of hypercalcemia of malignancy, acts on the common receptor for PTHrP and parathyroid hormone to enhance the renal retention of calcium and the release of calcium from bone, causing an increase in calcium in serum.³

In a study of 190 postmenopausal breast cancer cases and 172 controls, we reported that at the time of diagnosis, cases with newly-diagnosed, untreated breast cancer had significantly higher serum calcium levels.⁴ Although the results of that study are consistent with a calciotrophic effect of early breast cancer, we did not observe a significant association between breast cancer and tumor volume, which would be expected if the higher serum calcium was caused by the cancer. The lack of an association between serum calcium and tumor volume may have been due to the small number of cases. In this report we studied a large series of women with newly-diagnosed, untreated breast cancer (n = 555) in order to test the hypothesis that larger volume breast tumors are associated with higher serum calcium levels. In addition, we examined the relationship of serum calcium to other breast cancer clinical characteristics.

Results

Descriptive characteristics of the participants are shown in Table 1. Women with newly-diagnosed breast cancer had a mean (± SD) age of 61.8 ± 10.3 y and a BMI of 30.4 ± 6.8. Most women (95%) had early stage disease (defined as stages 0,I, IIA, IIB or IIIA) and a tumor grade of 1 or 2 (63%). Multiple linear regression showed that that age at diagnosis and tumor size were independent predictors of the corrected serum calcium for women overall. There was an effect for tumors that were estrogen receptor negative that was of borderline statistical significance (p = 0.066) (see Table 2). When women were stratified into pre -and post-menopausal status (n = 146, n = 343, respectively), age and BMI were the only significant independent predictor of corrected calcium among pre-menopausal women (p = 0.006). However, among women aged 56–79, BMI was not significant and tumor volume was the only significant predictor (p = 0.009).

Table 1. Descriptive characteristics of newly diagnosed breast patients aged 40- to 79-y-old

Variables corrected serum calcium levels (mg/dL)
	n = 555(%)	Mean (SD)	Range	P value
Age (years)				0.001
40–49	79 (14)	9.33 (0.42)	8.40–10.50
50–64	226 (41)	9.49 (0.45)	7.40–11.20
65–79	250 (45)	9.54 (0.48)	7.90–11.40
Body mass index (kg/m2)				0.58
Normal (< 25)	107 (19)	9.49 (0.46)	8.40–11.10
Overweight (25 - < 30)	195 (35)	9.46 (0.47)	7.40–11.40
Obese (> = 30)	253 (46)	9.52 (0.46)	8.20–11.20
Estrogen receptor				0.08
Negative	82 (16)	9.56 (0.43)	8.60–11.10
Positive	443 (84)	9.47 (0.47)	7.40–11.40
Progesterone receptor				0.64
Negative	133 (25)	9.50 (0.46)	8.40–11.20
Positive	394 (75)	9.48 (0.47)	7.40–11.40
Receptors combination				0.08
ER-/PR-	79 (14)	9.57 (0.43)	8.60–11.10
ER+/PR+	391 (70)	9.48 (0.47)	7.40–11.40
Tumor grades				0.14
Well-differentiated	94 (18)	9.42 (0.40)	8.40–10.80
Moderately differentiated	240 (45)	9.51 (0.45)	7.90–10.80
Poorly differentiated	196 (37)	9.53 (0.49)	8.40–11.40
Tumor size				0.04
≤ 2cm	399 (77)	9.46 (0.46)	7.40–11.10
> 2 cm	116 (23)	9.57 (0.46)	8.50–11.40
Stage at diagnosis				0.84
0	128 (24)	9.53 (0.50)	7.40–11.20
I	243 (45)	9.43 (0.46)	7.90–10.80
II*	121 (22)	9.56 (0.46)	8.50–11.40
III	39 (7)	9.47 (0.41)	8.80–10.40
IV	11 (2)	9.44 (0.43)	8.90–10.2

* Stage II vs. Stage I; p = 0.035

Table 2. Corrected serum calcium level determinants and coefficients for multiple linear regression analysis

Coefficients	(SE)	P value
Age at diagnosis	0.0053 (0.002)	0.0085
Body mass index (kg/m2)	0.0049 (0.003)	0.1031
Estrogen receptor negative	0.1540 (0.084)	0.0663
Progesterone receptor negative	-0.1145 (0.069)	0.1022
Poorly-differentiated tumors	0.0299 (0.050)	0.5505
Tumor size > 2 cm	0.1166 (0.051)	0.0233
Late stage (IIIB, IIIC, IV)	0.0077 (0.105)	0.9414

SE: standard error

We explored the relationship of serum albumin levels to tumor characteristics. Serum albumin levels were not associated with tumor size (p = 0.18) or with receptor status: ER-: 4.30 (2.5–5.0) vs. ER+: 4.30 (2.8–5.1); p = 0.34; Pr- 4.30 (2.5–5.0) vs. PR + 4.30 (2.8–5.1); p = 0.72. Tumor size was not significantly associated with receptor status, although both ER and PR negative tumors tended to be larger: ER vs. tumor size (p = 0.16); PR vs. tumor size (p = 0.07).

Discussion

In this study of 555 women aged 40–79 with newly-diagnosed, untreated breast cancer, we observed a significant positive association between tumor volume and serum calcium levels. The mean calcium levels for women with tumors ≤ 2 and > 2 cm was 9.46 (SD = 0.46) and 9.57 (SD = 0.46), respectively. This difference was statistically significant and remained so after adjustment for age, BMI, hormonal receptors, stage at diagnosis, and grade. Tumors that were estrogen receptor negative were associated with higher serum calcium levels, although the statistical ignificance of this finding was borderline (p = 0.07).

Hypercalcemia is a frequent complication of breast cancer, particularly in women with bony metastases.⁷ To our knowledge, this is the first report to demonstrate a positive association between serum calcium levels and tumor size in women with breast cancer who, with very few exceptions, were normocalcemic and had predominantly local disease. We did not observe a consistent relationship between serum calcium and tumor stage. However, tumor stage in breast cancer reflects both tumor volume and the pattern of metastasis. Because Stage I cancers are ≤ 2 cm, whereas Stage II includes cancers that are > 2 cm but < 5 cm, we compared serum calcium levels in Stages I and II. The mean serum calcium level was significantly higher in Stage II than in Stage 1 (9.56, SD = 0.46 vs. 9.43, SD = 0.46) (p = 0.035). Taken together with our previous finding that postmenopausal women with untreated breast cancer have higher serum calcium levels than postmenopausal women without breast cancer,⁴ our present finding that women with larger breast tumors have higher serum calcium levels than women with smaller tumors suggest that the higher levels of calcium in serum are a consequence of the cancer. (The alternate hypothesis, that higher serum calcium is a cause of breast cancer, does not specifically predict an association between serum calcium and tumor volume.) This hypothesis is plausible because approximately 60% of breast cancers express PTHrP, the causal agent of hypercalcemia of malignancy.⁸ PTHrP is also expressed in the normal and lactating breast, although at lower levels.⁹ It is noteworthy in this regard that the literature contains several case reports of hypercalcemia secondary to mammary hyperplasia (mammary gigantism).^10-12 The hyperplastic breast tissue has been shown to stain for PTHrP¹² and surgical reduction of the enlarged breasts caused a return to normocalcemia^10,12_. These observations support the hypothesis that the increase in number of PTHrP-secreting breast cells, whether benign or malignant, is responsible for the increase in serum calcium.

Stratification of the data into pre- and postmenopausal status indicated that the effect for serum calcium was found only among postmenopausal women (p = 0.009). The reasons for this difference are unclear, but could reflect the smaller sample size of pre-menopausal women (n = 146). Alternately, PTHrP expression has been shown to be higher in pre-menopausal women and is inversely correlated with age.¹³

Serum calcium levels and the risk of breast cancer have been examined prospectively in several cohorts. Almquist et al. reported that higher pre-diagnostic serum calcium levels were associated with a significantly increased risk of subsequent breast cancer among overweight and postmenopausal women in Sweden.¹⁴ A subsequent publication from this cohort reported higher serum calcium levels were positively associated with increased tumor aggressiveness in pre-menopausal women.¹⁵ Conversely, no association between either total or ionized serum calcium and breast cancer risk was found among a cohort of 2,762 U.S. women.¹⁶ Because our study was cross-sectional, it cannot determine conclusively whether the elevated calcium levels found in women with higher volume tumors are a cause of the larger tumor volume. Study of serum calcium levels before and after lumpectomy would be informative in resolving this issue.

ER- tumors were more likely to be > 2 cm than ER+ tumors (29% vs. 22%, p = 0.16). Similar findings were observed for PR (29% of PR- tumors were > 2cm, vs. 21% of PR+ tumors, p = 0.07). This finding is consistent with the findings of Chen et al., who reported that ER- breast tumors were significantly larger than ER+ tumors, as measured by MRI.¹⁷ In contrast to the findings of Al Murri et al., who reported that lower serum albumin levels (≤ 43 mg/dl) were associated with ER- and PR- tumors, we did not observe an association between serum albumin levels and receptor status.¹⁸

Our study has several limitations. The study was restricted to women with recent data on calcium and albumin. This raises the possibility of selection bias. However, for selection bias to have occurred, both the serum calcium levels and tumor characteristics would have to differ systematically between included and excluded women. Excluded women were significantly younger (mean age, 56, SD = 10, vs. 62, SD = 10). However, excluded women did not differ from included women on ER or PR status or on tumor grade. These similarities minimize concern about selection bias. Moreover, our findings on serum calcium levels in women with breast cancer are consistent with population-based studies of serum calcium among women without breast cancer.¹⁹ For example, our data demonstrate the expected positive association of serum calcium levels with increasing age (p < 0.001) and a (non-significant) trend for higher serum calcium among obese women. Other than age and BMI, which were controlled in this study, other factors have been reported to influence serum calcium levels in women, including alcohol use, early menarche and the use of hormonal therapy. The absence of data on these potential confounders is a limitation. However, the absolute difference in serum calcium levels associated with these factors is small.²⁰ Thus, in their population-based study of Norwegian women, Jorde et al. reported that influence of life style factors on serum calcium was minimal, with no factor accounting for > 1% of variation in total serum calcium levels.¹⁹

It is conceivable that our results could be influenced by differences in vitamin D. Although serum calcium levels may increase with serum levels of vitamin D, this relationship is most evident in the setting of vitamin D insufficiency (causing osteomalacia) and excess (causing hypercalcemia).²¹ However, serum calcium levels are tightly controlled and are little influenced by vitamin D at normal vitamin D levels. For example, in a cross-sectional study of 918 postmenopausal women, Need et al. reported that the Pearson correlation coefficient for serum 25-Hydroxyvitamin D and ionized calcium was 0.02.²² Alternately, serum levels of 1,25-Dihydroxyvitamin D have been reported to decline with increasing tumor stage in breast cancer.²³ If this were true in our study, it would act to decrease the serum calcium levels in women with larger tumors (i.e., the effect would be conservative).

We did not abstract data on race. Because the overwhelming majority (> 95%) of women served by the Sanford Health catchment area is Caucasian, any bias caused by the inclusion of non-Caucasian women would be minimal. Finally, HER-2/neu status was available for only a small number of women and thus could not be analyzed meaningfully. Conversely, the study has several strengths: the sample size is large and the cross-sectional design captures women before they received treatment for their cancer. In addition, women with conditions that could confound serum calcium levels (e.g., chronic kidney disease) were excluded. Lastly, the data were abstracted by individuals unaware of the study hypotheses.

In summary, in this cross-sectional study of 555 women with untreated breast cancer, serum calcium levels at diagnosis, adjusted for covariates, were significantly higher in women with breast tumors > 2 cm. A plausible explanation for this observation is that the higher serum calcium is due to the effects of a tumor-derived factor(s) that elevates serum calcium. If these data are confirmed by subsequent investigations, tracking serum calcium levels over time may be useful in the setting of breast cancer screening. In healthy individuals, serum calcium levels are very stable, highly reproducible, and (except at extreme intakes) are little affected by diet.²⁴ Serum calcium levels increase slightly after menopause.¹⁹ Thus, if serum calcium levels obtained during annual physical exams in individual post-menopausal women were compared with one another, we speculate that an increase in serum calcium might increase suspicion of either primary hyperparathyroidism or of an occult breast tumor.²⁵ This hypothesis could be tested in the screening setting by comparing the risk of breast cancer in women who show a positive change in annual calcium measurements to the risk in women whose serum calcium levels do not.

Methods

This is a cross-sectional study of clinical characteristics of newly diagnosed breast cancer using medical records of patients diagnosed between January 2003 and December 2010. Patients were identified from the electronic medical database of Sanford Health, North Dakota, USA, which serves the Metropolitan area of Fargo, ND. This study was approved by the Institutional Review Boards of Sanford Health and the University of North Dakota.

The inclusion criteria were women of screening age for breast cancer, aged 40–79, with new histologically confirmed epithelial breast cancer diagnosed between 2003 and 2010. The exclusion criteria included any cancer other than primary breast cancer, recurrence of breast cancer, diagnosis outside the study period, and missing calcium level within a year prior to diagnosis. Additionally, we excluded women with medical conditions known to involve a dysregulation of serum calcium, e.g., a diagnosis of hyperparathyroidism or a diagnosis of chronic kidney disease, and/or an estimated glomerular filtration rate (eGFR) < 60 ml/min.

Our study sample began with 1900 women with a breast cancer diagnosis provided by the Sanford Department of Quality Improvement. Exclusions were as follows: outside the age range (103 cases: 14 women younger than 40 y and 89 older than 79 y); missing data for calcium in the study period (293); missing dates for serum calcium (753); missing data for body mass index (BMI) [weight (kg)/height (m²)] (91); recurrence of cancer (49); chronic kidney disease (49), and non-epithelial cancers of breast, e.g., sarcomas and lymphomas (7). After these exclusions, 555 women were eligible for study.

The serum calcium level was the level reported within one year prior to the date of breast cancer diagnosis median (range): 58 (0–365) days. Approximately half of total serum calcium is in the free or unbound state, which is the biologically active fraction of total serum calcium. The remaining portion is bound to albumin and to anions such as lactate and phosphate. We used the formula for “corrected” serum calcium from using the formula from the Modification of Diet in Renal Disease, which adjusts the total serum calcium for variations in calcium level caused by binding to albumin.⁵ If the serum albumin level was < 4 g/dL, the total serum calcium was adjusted by a standard formula [(corrected serum calcium = total calcium+0.8*(4-albumin)]. If the serum albumin was ≥ 4 g/dL or if the albumin was not measured, then the total serum calcium was used. Furthermore, if serum albumin was < 4 g/dL but the difference between the dates of calcium and albumin measurement was > 10 d, total serum calcium was used. We adopted the “10 day” rule because the half-life of albumin in serum is ~21 d.⁶ The majority of women (403, 72%) had normal serum albumin levels and no correction was needed. Serum albumin levels were corrected for 28 women (5%). Fifteen women (3%) had abnormal albumin levels, but the difference in the dates of calcium and albumin measurements were beyond the 10 d window; thus unadjusted serum calcium levels were used for these women. Among these women, 7/15 (47%) had albumin values of 3.9 g/dL, 4 (27%) were between 3.6–3.8 g/dL, and 4 were < 3.8 g/dL (27%). Serum albumin levels were missing in 109 women (20%).

We grouped BMI into < 25 and ≥ 25. Tumor grades were obtained from the histopathology report and graded as 1 (well-differentiated), 2 (moderately differentiated) and 3 (poorly differentiated). Tumor stage was classified according to the American Joint Committee on Cancer TNM staging system. Tumor size was obtained from the pathology reported and classified as ≤ 2 and > 2cm.

Hormone receptor status was reported in the histopathology report as positive or negative for estrogen receptor (ER) or progesterone receptor (PR). Hormone receptor assay was performed by immunoperoxidase staining for ER, PR on paraffin-embedded sections of formalin-fixed tissue from tumor. The method employed the Ventana Medical Systems antibody clones using 3,3′-diaminobenzidine tetrahydrocholoride DAB detection system. Scoring for ER and PR is based on percent of staining of tumor cell nuclei with the following determinations: negative (0%), low positive (< 10%), and positive (> 10%).

We analyzed data on age at diagnosis (in years), serum calcium (mg/dL), serum albumin, BMI, breast cancer stage, primary tumor size, histopathological grade of the cancer, and hormone receptors status. We used age 55 as a cutpoint to approximate menopausal status and performed a stratified analysis for women who were pre- (aged 55 or younger) and post-menopausal (aged 56 and older).

Statistical analyses

Descriptive statistics are presented as percentages and means with standard deviations (SD) and ranges. Chi-square tests were used to analyze the differences between groups of categorical variables and Wilcoxon signed-rank tests for continuous variables. We used multiple linear regression to assess significant predictors of corrected serum calcium (outcome). We evaluated the following variables as potential predictors: age at diagnosis, BMI, estrogen receptor status, progesterone receptor status, stage of disease, tumor grade, and tumor size. We also examined the correlations between serum albumin, hormonal receptors, and tumor size. All p values are two-sided with p value < 0.05 considered significant. Analyses used SAS software V 9.2 (SAS Institute).

Acknowledgments

We thank Ms. Mary Kara, Department of Quality Improvement, Sanford Health, for assistance with the electronic medical records data.

Disclosure of Potential Conflicts of Interest

The authors have no financial or other conflicts of interest