Abstract
Background: Higher levels of total and ionized serum calcium have been shown to predict fatal prostate cancer in prospective studies. Because the follow-up time in these studies was relatively short, these associations could reflect the effect of clinically significant but occult prostate tumors on serum calcium levels. If this were true, prostate cancer mortality rates among men with higher levels of serum calcium should be higher during the early follow-up period and should decline thereafter.
Methods: We tested this hypothesis by estimating the relative risk of death from prostate cancer in the National Health and Nutrition Examination Survey III for incremental increases in total and ionized serum calcium using Cox proportional hazards regression with time-dependent effects.
Results: Forty-nine (49) fatal prostate cancers occurred over 204 months of follow-up and 1,069,327 person–months of observation. Men with higher total serum calcium and higher serum ionized calcium had increased risks of fatal prostate cancer during the first 96 months of follow-up [Relative Hazard (RH) = 1.50 per 0.1 mmol/L total serum calcium, 95% confidence interval (CI) = 1.04–2.17; RH = 1.72 per 0.05 mmol/L ionized calcium, 95% CI = 1.11–2.66]. Evidence of an association between total and ionized serum calcium and prostate cancer deaths was not significant after 96 months.
Conclusions: Our analyses support the hypothesis that the elevated risk for fatal prostate cancer observed in men with high serum calcium is because of the presence of extant, but occult prostate cancer.
Impact: These findings have implications for the potential use of serum calcium in the detection of clinically significant prostate cancer. Cancer Epidemiol Biomarkers Prev; 21(10); 1768–73. ©2012 AACR.
Introduction
The subject of calcium and the risk of prostate cancer began with ecologic studies that showed positive correlations between international mortality rates from prostate cancer and the per capita consumption of dairy products (1). These studies were followed by investigations of prostate cancer risk in individuals in relation to dairy product and calcium consumption (e.g., refs. 2–4). The consensus from the analytic epidemiologic studies is that high dietary calcium intake increases the risk of prostate cancer, especially the risk of advanced and/or fatal cancer. Thus, the Agency for Health Care Research and Quality noted that 3 of the 4 cohort studies rated “A” for methodologic quality show significantly increased risk for prostate cancer with diets high in calcium (5). Similarly, the World Cancer Research Fund/American Institute for Cancer Research concluded that foods containing calcium are a probable cause of prostate cancer (6).
We examined the relationship between calcium in serum and prostate cancer in 2 prospective cohorts, the National Health and Nutrition Examination Survey NHANES and NHANES III. In the first NHANES and the NHANES Epidemiologic Follow-up Study, 85 incident prostate cancers and 25 prostate cancer deaths occurred over 46,188 person–years of follow-up. Serum calcium was measured an average of 9.9 years before the diagnosis of prostate cancer. When men in the top tertile of total serum calcium were compared with men in the bottom tertile, the relative hazard for incident prostate cancer was not significantly elevated [RH = 1.31, 95% confidence interval (CI) = 0.77–2.20]. Conversely, the multivariable-adjusted relative risk for fatal prostate cancers was 2.68 (95% CI = 1.02–6.99; P trend = 0.04; 7). We sought to confirm the association between serum calcium and risk of fatal prostate cancer in NHANES III and the linked mortality file, which included data on ionized serum calcium, the biologically active fraction of total serum calcium. Twenty-five prostate cancer deaths occurred over 56,625 person–years. Serum calcium was measured an average of 5.3 years before death from prostate cancer. The multivariate-adjusted relative risks for death from prostate cancer for men in the highest tertile of total serum calcium and of ionized serum calcium were 2.07 (95% CI = 1.06–4.04) and 3.18 (95% CI = 1.09–9.28), respectively (8). These data confirm that high serum calcium levels predict death from prostate cancer.
In contrast, total serum calcium was not predictive of prostate cancer in a prospective study in Sweden (9). Because the Swedish study combined incident and fatal cases, the risk may have been “diluted” by the incident cancers. However, these studies also differed in the length of follow-up: the average follow-up was >22 years in the Swedish study versus <10 years in the United States studies. Thus, it is possible that the elevated risks that we observed in NHANES were because of the presence of extant, but undetected prostate cancer. We reasoned that if this were true, prostate cancer mortality rates among men with higher serum calcium should be higher in the initial follow-up period and should decline thereafter. We tested this hypothesis in the expanded 2010 release of the public-use mortality linkage file of NHANES III that includes deaths recorded through December 31, 2006.
Materials and Methods
We used baseline data collected between 1988 and 1994 for the NHANES III and follow-up information through 2006 in the “2010 release” of the public-use linked mortality file to compute survival from time of blood draw to death from prostate cancer for up to 204 months (17 years) of follow-up. Men who reported a prior personal history of cancer were excluded. Deaths from all other causes were censored at the time of death. We used Cox proportional hazards regression accounting for sample weights and the complex sample design to estimate the relative risk of death from prostate cancer for incremental increases in total serum calcium and ionized serum calcium. We examined potential confounding by age, body mass index (BMI), race/ethnicity, serum 25-hydroxyvitamin D (25-OHD), and serum albumin levels. To account for the time dependence of the relationships, separate Cox models were fit for early follow-up (events occurring 0–96 months after baseline; n = 24) and late follow-up (97–204 months from baseline; n = 25) based on the median time to death from prostate cancer (94 months) rounded to the next whole year. We computed P values for interactions between baseline calcium concentration and follow-up time by fitting a Cox model with an interaction term between calcium as a continuous value and time period. Statistical analyses were carried out using R v. 2.14.2 with the “survival” package (10).
Both total and ionized serum calcium levels were measured using ion-specific electrodes (11). Approximately half of total serum calcium is in the “free” or ionized state. Approximately 40% is bound to serum proteins, principally albumin, and the remainder is bound to anions such as phosphate. Because the binding of calcium to proteins is affected by pH, ionized calcium in blood is commonly corrected to a standard pH (12). The ionized calcium was pH adjusted.
Results
There were 6,707 men 18 years of age or older. Forty-nine (49) fatal prostate cancers occurred over 1,069,327 person–months of observation (Table 1). After adjusting for age, BMI, and serum albumin levels, men with higher total serum calcium had a significantly increased risk of fatal prostate cancer over the first 96 months of follow-up (Table 2). The relative hazard for prostate cancer death was 1.50 (95% CI = 1.04–2.17) for each 0.1 mmol/L increase in total serum calcium. Similarly, men with higher serum ionized calcium at baseline had a significantly higher relative hazard of prostate cancer death over the first 96 months (RH = 1.72 per 0.05 mmol/L of ionized calcium, 95% CI = 1.11–2.66). In contrast, for deaths that occurred more than 96 months after baseline, there was no significant association between total or ionized calcium and risk for prostate cancer mortality in multivariable-adjusted models. The P value for interaction between baseline calcium and follow-up time was 0.049 for ionized calcium and 0.058 for total serum calcium.
Selected characteristics of men in the NHANES III by tertile of serum total calcium concentration
| . | Baseline total serum calcium tertile . | ||
|---|---|---|---|
| . | Tertile 1 . | Tertile 2 . | Tertile 3 . |
| Number of participants | 5,033 | 4,510 | 4,669 |
| Weighted population | 53,239,124 | 50,734,316 | 58,638,470 |
| Prostate cancer deaths through 2006 | 22 | 10 | 17 |
| Person–months of follow-up | 805,945 | 730,711 | 760,416 |
| Mean total calcium (mmol/L) | 2.2 (0.09) | 2.3 (0.02) | 2.4 (0.07) |
| Mean ionized calcium (mmol/L) | 1.2 (0.04) | 1.2 (0.03) | 1.3 (0.04) |
| Mean Age (y) | 47.7 (16.96) | 43.0 (15.92) | 38.2 (15.40) |
| Mean BMI (kg/m2) | 26.7 (4.87) | 27.0 (5.23) | 26.2 (4.48) |
| Mean 25-OHD (ng/mL) | 30.0 (10.91) | 30.9 (11.46) | 32.2 (11.82) |
| Mean Albumin (g/dL) | 4.2 (0.34) | 4.3 (0.31) | 4.4 (0.31) |
| . | Baseline total serum calcium tertile . | ||
|---|---|---|---|
| . | Tertile 1 . | Tertile 2 . | Tertile 3 . |
| Number of participants | 5,033 | 4,510 | 4,669 |
| Weighted population | 53,239,124 | 50,734,316 | 58,638,470 |
| Prostate cancer deaths through 2006 | 22 | 10 | 17 |
| Person–months of follow-up | 805,945 | 730,711 | 760,416 |
| Mean total calcium (mmol/L) | 2.2 (0.09) | 2.3 (0.02) | 2.4 (0.07) |
| Mean ionized calcium (mmol/L) | 1.2 (0.04) | 1.2 (0.03) | 1.3 (0.04) |
| Mean Age (y) | 47.7 (16.96) | 43.0 (15.92) | 38.2 (15.40) |
| Mean BMI (kg/m2) | 26.7 (4.87) | 27.0 (5.23) | 26.2 (4.48) |
| Mean 25-OHD (ng/mL) | 30.0 (10.91) | 30.9 (11.46) | 32.2 (11.82) |
| Mean Albumin (g/dL) | 4.2 (0.34) | 4.3 (0.31) | 4.4 (0.31) |
NOTE: Where means are given, SD appear in parentheses. Means and SD account for sample weights and complex sampling design of NHANES III. The approximate normal range for total serum calcium is 2.17–2.52 mmol/L (8.7–10.1 mg/dL) and for ionized serum calcium is 1.12–1.32 mmol/L (4.6–5.3 mg/dL).
Multivariable-adjusted relative hazards of time-dependent prostate cancer mortality in relation to total and ionized serum calcium concentrations in men in the NHANES III with mortality linkage through 2006
| . | Deaths 0–8 years . | Deaths 8+ years . | ||
|---|---|---|---|---|
| Characteristic . | Relative hazard . | 95% CI . | Relative hazard . | 95% CI . |
| Ionized calcium (per 0.05 mmol/L) | 1.73 | (1.13–2.63) | 0.82 | (0.35–1.95) |
| Ionized calcium tertile 1 | 1.00 | Reference | 1.00 | Reference |
| Tertile 2 | 1.69 | (0.29–9.96) | 0.44 | (0.10–1.95) |
| Tertile 3 | 4.46 | (0.89–22.38) | 1.01 | (0.23–4.43) |
| Age (per year) | 1.13 | (1.10–1.17) | 1.14 | (1.10–1.17) |
| BMI (per kg/m2) | 0.86 | (0.73–1.01) | 1.06 | (0.98–1.15) |
| Serum albumin (per g/dL) | 0.52 | (0.09–3.19) | 0.29 | (0.07–1.21) |
| 25-OH-vitamin D (per ng/mL) | 0.97 | (0.93–1.01) | 1.02 | (0.98–1.07) |
| Total calcium (per 0.1 mmol/L) | 1.49 | (1.04–2.14) | 0.75 | (0.57–0.99) |
| Total calcium tertile 1 | 1.00 | Reference | 1.00 | Reference |
| Tertile 2 | 0.82 | (0.18–3.81) | 0.69 | (0.19–2.51) |
| Tertile 3 | 1.76 | (0.46–6.78) | 1.08 | (0.26–4.41) |
| Age (per year) | 1.13 | (1.10–1.17) | 1.14 | (1.10–1.18) |
| BMI (per kg/m2) | 0.85 | (0.72–1.00) | 1.06 | (0.98–1.15) |
| Serum albumin (per g/dL) | 0.39 | (0.05–3.12) | 0.39 | (0.10–1.45) |
| 25-OH-vitamin D | 0.97 | (0.93–1.02) | 1.02 | (0.98–1.07) |
| . | Deaths 0–8 years . | Deaths 8+ years . | ||
|---|---|---|---|---|
| Characteristic . | Relative hazard . | 95% CI . | Relative hazard . | 95% CI . |
| Ionized calcium (per 0.05 mmol/L) | 1.73 | (1.13–2.63) | 0.82 | (0.35–1.95) |
| Ionized calcium tertile 1 | 1.00 | Reference | 1.00 | Reference |
| Tertile 2 | 1.69 | (0.29–9.96) | 0.44 | (0.10–1.95) |
| Tertile 3 | 4.46 | (0.89–22.38) | 1.01 | (0.23–4.43) |
| Age (per year) | 1.13 | (1.10–1.17) | 1.14 | (1.10–1.17) |
| BMI (per kg/m2) | 0.86 | (0.73–1.01) | 1.06 | (0.98–1.15) |
| Serum albumin (per g/dL) | 0.52 | (0.09–3.19) | 0.29 | (0.07–1.21) |
| 25-OH-vitamin D (per ng/mL) | 0.97 | (0.93–1.01) | 1.02 | (0.98–1.07) |
| Total calcium (per 0.1 mmol/L) | 1.49 | (1.04–2.14) | 0.75 | (0.57–0.99) |
| Total calcium tertile 1 | 1.00 | Reference | 1.00 | Reference |
| Tertile 2 | 0.82 | (0.18–3.81) | 0.69 | (0.19–2.51) |
| Tertile 3 | 1.76 | (0.46–6.78) | 1.08 | (0.26–4.41) |
| Age (per year) | 1.13 | (1.10–1.17) | 1.14 | (1.10–1.18) |
| BMI (per kg/m2) | 0.85 | (0.72–1.00) | 1.06 | (0.98–1.15) |
| Serum albumin (per g/dL) | 0.39 | (0.05–3.12) | 0.39 | (0.10–1.45) |
| 25-OH-vitamin D | 0.97 | (0.93–1.02) | 1.02 | (0.98–1.07) |
NOTE: Models are adjusted for age and BMI, serum albumin, and serum 25-OHD and account for survey weights and the complex sampling design of NHANES III.
Discussion
In this prospective, nationally representative sample of men without clinical prostate cancer, men with higher total serum calcium and/or higher ionized serum calcium had significantly elevated risks of fatal prostate cancer over the first 96 months of follow-up. For prostate cancer deaths more than 96 months after baseline, there was no evidence of a positive association. These data are consistent with our previous findings that higher serum calcium levels were associated with an increased risk of fatal prostate cancer for follow-up less than 10 years. This is the first demonstration that the risk of fatal prostate cancer associated with serum calcium is dependent upon the duration of follow-up.
Our findings that higher serum calcium levels predict fatal prostate cancer within a relatively short follow-up period are consistent with (at least) 2 hypotheses, which we name the “host” and the “tumor” hypotheses. These hypotheses differ about the cause of the elevated serum calcium. The host hypothesis suggests that men who have higher serum calcium levels develop prostate cancers that are rapidly fatal, i.e., the higher serum calcium is a characteristic of the host. The tumor hypothesis suggests that prostate tumors that are potentially fatal cause higher calcium levels in serum, i.e., the higher serum calcium is a remote effect of the tumor (13). The host and tumor hypotheses make different predictions about the time to death. The host hypothesis predicts an increased risk of death among men with higher serum calcium, but does not specify when this increase should occur (14). Conversely, the tumor hypothesis predicts that the increased death rate should occur relatively early because the elevations in serum calcium are caused by clinically significant tumors that are extant. Our finding that death rates were elevated in the first 8 years of follow-up is consistent with the predictions of the tumor hypothesis. This interpretation also is consistent with the duration of preclinical prostate cancer, which is estimated to be 11 to 12 years in Caucasians (15).
In a recent prospective study in Sweden (the AMORIS study) with a mean follow-up of 12 years, Van Hemelrijck and colleagues reported no significant association between serum calcium and fatal prostate cancer (N = 731 fatal cancers). However, there was a significant positive association between albumin-adjusted serum calcium and the risk of “aggressive prostate cancers” (N = 579), defined as fatal cancers that occurred within 2 years following diagnosis (16). These findings are consistent with our conclusion of a time-dependent association between serum calcium and the risk of fatal prostate cancer.
The tumor hypothesis is consonant with the molecular endocrinology of prostate cancer. For example, relative to benign prostate disease, the expression of parathyroid hormone-related protein (PTHrP) is significantly increased in prostate cancer and has been described in 93% of specimens of localized prostate cancer (17). PTHrP raises calcium levels in serum and is the principal agent of hypercalcemia of malignancy (18). PTHrP also promotes the metastasis of prostate cancer cells to bone (19, 20). Thus, we suggest that the production of PTHrP by local prostate tumors is a plausible mechanism that could underlie the association between higher serum calcium and decreased prostate cancer survival. Hypercalcemia is known to predict decreased survival in other cancers that express PTHrP, including cancers of the breast and lung (21, 22).
Although PTHrP expression in localized prostate cancer is common, hypercalcemia in advanced prostate cancer is rare, accounting for less than 2% of published cases (23). Men with metastatic prostate cancer often develop low normocalcemia or hypocalcemia because blastic metastases in bone (the site of the overwhelming majority of metastases in prostate cancer) trap serum calcium into the bony lesions (24). The hypocalcemic effect of the blastic lesions likely overwhelms the small elevations in serum calcium that may be caused by the organ-confined lesions, which occur earlier in the natural history of prostate cancer (25). This interpretation is consistent with the findings of Dai and colleagues, who reported that, at the time of prostate biopsy, men with newly diagnosed prostate cancer (who presumably had only localized cancer) had significantly higher total serum calcium levels than men with benign prostatic hyperplasia (26).
Our findings should be evaluated in light of several potential limitations. For example, the use of total serum calcium is subject to misclassification. Approximately 50% of total serum calcium is bound to serum proteins and anions and thus its concentration may be influenced by serum levels of these factors. It is noteworthy that the association between ionized serum calcium and fatal prostate cancer was stronger than that for total serum calcium, with a RH of 1.72 per 0.05 mmol/L increase in ionized calcium versus a RH of 1.50 per 0.1 mmol/L increase in total calcium, which is consistent with the fact that ionized calcium is the biologically active fraction of total serum calcium. Serum albumin was an important confounder of the association between total serum calcium and prostate cancer mortality, particularly in the later follow-up period. We observed a nonsignificant relative hazard for prostate cancer mortality of 2.56 per 1 g/dL decrease in albumin concentration (P = 0.14) for deaths more than 96 months after baseline. Adjustment of total serum calcium for serum albumin strongly attenuated an apparent inverse association in an unadjusted model between total serum calcium and prostate cancer mortality (RH = 0.73; P = 0.051). We cannot rule out residual confounding by albumin as an explanation for the paradoxical inverse association between total serum calcium and prostate cancer mortality for deaths that occurred more than 96 months after baseline. Conversely, it is possible that low serum albumin may have a biologic role in prostate cancer. Albumin levels are known to decrease as part of a systemic inflammatory response in cancer (27). Moreover, low albumin levels within the normal range have been reported to be significantly associated with an increased risk of local metastatic disease (28) and with death from prostate cancer (29).
It is possible that our results for serum calcium could be influenced by other factors, e.g., dietary calcium and serum levels of vitamin D. Serum levels of calcium are generally under tight homeostatic control and are not altered by calcium intake within the normal range of calcium intake (30). Although the consumption of calcium supplements may transiently cause higher ionized calcium levels in serum, ionized calcium levels return to baseline within several hours and total serum calcium is little affected (31). The fact that both ionized and serum levels were elevated in men who developed fatal cancer suggests that these elevations are unlikely to be due solely to the use of calcium supplements. In theory, it is possible that the serum calcium levels could be confounded by serum levels of 25-OHD. Our results are unlikely to be influenced by vitamin D for several reasons. First, we previously examined the relationship between serum 25-OHD and total serum calcium in NHANES 2005 and reported that the correlation between these variables was modest (r = 0.14, N = 1,273; 32). Moreover, adjustment of these data for serum levels of 25-OHD slightly increased the strength of our findings for serum calcium. Thus, we conclude that serum levels of 25-OHD did not exert an important influence on these results. Finally, our results may have been influenced by the inclusion of occult, metastatic cancer. However, because metastatic prostate cancer is characterized by low normocalcemia and/or hypocalcemia, this would bias our results in a conservative direction (33).
This study has several strengths: it is prospective, population-based, uses a population that is representative of the United States population, and includes >1 million person–months of observation. The use of a standardized laboratory in NHANES minimizes the effects of potential laboratory biases. Furthermore, the use of ionized serum calcium as an exposure variable adds to the biologic plausibility of our findings. Ionized serum calcium is the molecule that interacts directly with cellular receptors, e.g., the calcium sensing receptor and calcium-dependent voltage gated channels that are expressed on prostate cancer cells (34, 35).
Our findings raise an apparent paradox in the current understanding of the role of calcium in prostate cancer. Diets that are high in calcium are an established risk factor for prostate cancer and men with genotypes that are associated with greater calcium absorption also have an increased risk (36). In concert with our previous findings of an increased risk for high calcium in serum, it is tempting to interpret these data to indicate that high serum calcium is the cause of the increased prostate cancer risk. However, our present results suggest that the higher serum calcium is, at least in part, a consequence of the cancer rather than its cause. We suggest that this paradox can be resolved by the recognition that higher ionized calcium levels in serum (e.g., from diet) stimulate prostate cancer cells to produce PTHrP, which in turn raises calcium in serum in a positive feedback loop (37). Thus, we speculate that dietary calcium intake and increased serum calcium resulting from localized prostate cancer act to increase the risk of fatal prostate cancer via the same mechanism—an increase in ionized serum calcium.
Our findings that calcium levels significantly predict fatal prostate cancer in the near-term suggest that calcium measurements may help inform clinical decision making (38). For example, most men with a positive screening prostate-specific antigen (PSA) test that is between 4 and 10 ng/mL who are referred for a prostate biopsy do not have prostate cancer (39). Thus, most of these biopsies are conducted unnecessarily. Moreover, PSA cannot distinguish life threatening from clinically insignificant prostate tumors. This is especially problematic because the prevalence of clinically insignificant prostate cancers among older men is high. Unnecessary biopsies may detect these insignificant lesions and lead to overtreatment of prostate cancer (40, 41). Our data indicate that total and ionized serum calcium levels are significantly elevated in men without clinical prostate cancer who die of prostate cancer within 8 years. Thus, higher serum calcium may be a marker of prevalent, clinically significant prostate cancer that could aid in its early detection. Studies of the test characteristics of serum calcium in the diagnostic setting, e.g., in association with serum PSA, should be an important area for future epidemiologic research.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: G.G. Schwartz, H.G. Skinner
Development of methodology: G.G. Schwartz, H.G. Skinner
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.G. Skinner
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G.G. Schwartz, H.G. Skinner
Writing, review, and/or revision of the manuscript: G.G. Schwartz, H.G. Skinner
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.G. Skinner
Study supervision: H.G. Skinner
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.
RSS Feeds