Randomized Study of High-Dose Pulse Calcitriol or Placebo prior to Radical Prostatectomy

Population studies suggest that environmental factors contribute significantly to the risk of developing prostate cancer (1, 2). It may be possible to identify interventions that reduce the risk of this neoplasm. Currently, a large randomized trial is evaluating selenium and vitamin E for prostate cancer prevention (3).

Conduct of prevention trials require enormous resources because of the large numbers of patients and the years of follow-up needed to achieve sufficient statistical power. Because such studies frequently take more than a decade to complete, agents chosen for such trials should be carefully vetted. Whereas preclinical models of prostate cancer may be helpful in identifying potential chemoprevention agents, the use of prostate cancer animal models for this purpose has not been fully validated. Reliable approaches for the selection of chemoprevention candidates are therefore needed.

Examination of candidate prevention agents using biomarker end points has been proposed as a method for vetting agents before testing in large clinical trials (4, 5). Treatment of patients with candidate agents prior to scheduled biopsy or surgical resection of malignancy allows direct examination of the effect of the agent on tumor tissue. In this study, we selected this approach to test several hypotheses about the effect of calcitriol (1,25-dihydroxycholecalciferol), the active form of vitamin D, on human prostate cancer.

Substantial preclinical evidence supports testing vitamin D receptor (VDR) ligands for prostate cancer treatment and prevention. In prostate cancer, activity has been reported both in vitro (6-12) and in animal models (8, 13). These models have suggested several mechanisms of action, including inhibition of invasion, cell cycle arrest, induction of apoptosis, and inhibition of angiogenesis. In addition, a VDR ligand has been shown to prevent the development of carcinoma of the prostate and seminal vesicle in the Lobund-Wistar rat model (14).

Antineoplastic activity of calcitriol in preclinical systems is seen at concentrations significantly above the physiologic range, typically above 1 nmol/L. Such concentrations cannot be reached with conventional daily dosing of calcitriol due to predictable hypercalcemia and hypercalcuria (15). However, they are safely achievable with a weekly dosing schedule (16). When tested in a phase II trial carried out in hormone-naive patients with a rising serum prostate-specific antigen, weekly calcitriol produced reductions in serum prostate-specific antigen in 14% of patients and was associated with an increase in the prostate-specific antigen doubling time (17). In this study, we sought to determine the effect of weekly calcitriol compared with placebo on several relevant biomarkers in human prostate cancer. We measured the expression level of the VDR as well as molecules involved with cell cycle regulation and proliferation, including proliferating cell nuclear antigen (PCNA; refs. 18-21), PTEN (MMAC1/TEP1; refs. 22-24), c-Myc (25-28), and transforming growth factor (TGF) β receptor type II (TGFβ RII) (29-34) and regulators of apoptosis, including Bcl-2 (12, 35-39) and PTEN in patients treated with either calcitriol or placebo prior to prostatectomy.

Patients with histologically proven adenocarcinoma of the prostate who had selected prostatectomy as primary therapy were considered eligible if they met the following criteria: age ≥18, Eastern Cooperative Oncology Group performance status ≤2, serum creatinine ≤1.3 mg/dL, serum calcium ≤10.5 mg/dL, serum phosphate ≤4.2 mg/dL, hemoglobin >10 g/dL, total bilirubin <1.2 mg/dL, and signed informed patient consent. Patients were excluded for: other preoperative or prior treatment directed at prostate cancer, significant active medical illness, history of hypercalcemia, kidney stones within 5 years, heart failure or significant heart disease including myocardial infarction in the last 3 months, known cardiac ejection fraction <30%, current digoxin therapy, thiazide diuretic therapy within 7 days, and unwillingness or inability to stop all magnesium-containing antacids, bile resin-binding drugs, or calcium supplements for the duration of the study. Written informed consent was obtained from all patients. The protocol was approved by the Institutional Review Boards of the Oregon Health & Science University and the Portland VA Medical Center.

Patients were asked to maintain a reduced calcium diet as previously described (16). Each study drug capsule consisted of two calcitriol (Rocaltrol, Roche Pharmaceuticals, Nutley, NJ) 0.5-μg capsules placed inside a brown, gelatin capsule or the same gelatin capsule filled with starch (placebo). The study drug was administered at a dose of 1 capsule per 2 kg body weight (0.5 μg/kg of calcitriol in the active treatment arm) and was given p.o. once a week for a 4-week period ending 3 to 4 days prior to prostatectomy. Each weekly dose was divided equally into 4 doses and taken p.o. during each hour for 4 hours. Up to an additional 4 weeks of treatment were allowed if prostatectomy was delayed for unrelated reasons.

The screening evaluation included a complete history and physical exam, hemoglobin, serum creatinine, serum calcium, serum phosphate, total serum bilirubin and prostate-specific antigen. In addition, serum calcium was checked on the third day of the second week of treatment. Complete blood count, serum calcium, phosphate, creatinine, total bilirubin, and prostate-specific antigen were checked at the end of treatment. All toxicities were graded according to National Cancer Institute Common Toxicity Criteria (version 2.0).

Prostatectomy specimens were weighed, measured, inked with two different colors to distinguish the left and right lobes, and then fixed in 70% ethanol for a minimum of 48 hours. Fixed specimens were submitted in toto for paraffin embedding and sectioning as whole-mounted transverse sections. All sections from each case were stained with H&E and reviewed by the same pathologist who was blinded to treatment assignment. A single, experienced pathologist was chosen for consistency of interpretation. The carcinomas were graded according to Gleason criteria. Four representative areas from each whole-mount paraffin block were cut out and placed in standard-sized OmniSette tissue cassettes (Fisher Scientific, Pittsburgh, PA). These areas were chosen as the best composite representation of tumor based on H&E morphology. These were re-embedded, sectioned at 4 μm, and mounted on poly-l-lysine–coated slides for immunohistochemistry. Antibodies and antigen retrieval methods are summarized in Table 1.

All antibodies were diluted in TBS/0.3% bovine serum albumin. Sections of paraffin embedded tissue were deparaffinized and rehydrated in a Leica Autostainer XL (Bannockburn, IL). Antigen-retrieval methods are described in Table 1. With the exception of the PCNA sections, all slides were stained using an automated immunostainer (model DC3400, Dako, Carpinteria, CA) under room temperature. Following preincubation with blocking buffer (TBS/0.3% bovine serum albumin/0.05% Tween 20) primary antibody was added for 45 minutes. The remainder of the staining protocol was exactly as described in Renshaw et al. (40).

After deparaffinization and rehydration, sections for PCNA staining were postfixed in Z-fix (Anatech LTD, Battle Creek, MI) for 30 minutes at room temperature, rinsed with deionized water, and subjected to antigen-retrieval steps as described in Table 1. Incubation with the primary antibody was done overnight (approximately 16 hours) at 4°C in a humidified staining chamber. The remainder of the staining protocol was the same as for the other antibodies, except the steps were done by hand rather than on an autostainer. For all of the antibodies used in the study the staining was visualized by incubating the slides for 10 minutes with 3,3′-diaminobenzidine solution (K3466, Dako), after which they were rinsed in water, counterstained with hematoxylin, dehydrated in a Leica Autostainer XL, coverslipped, and reviewed.

For biomarkers that were measured as a percentage of positively staining cells (or nuclei, where appropriate, indicated in Table 5), 1,000 cells from the most representative blocks of tumor were counted to ensure adequate confidence intervals by a single experienced pathologist who was blinded to treatment assignment. The 95% confidence intervals (95% CI) were calculated by the formula:

Sample size was calculated based on published proliferation biomarker levels [% PCNA, 11.84; SD, 2.9 (ref. 18) and % MiB-1, 8.15; SD, 4.07 (ref. 41)] and variability in prostate cancer with the goal of having adequate power to detect a 50% difference in means of each biomarker between the treated and placebo group using a two-group, two-sided t test with a 0.050 significance level. A total enrollment of 17 patients per arm was planned with enrollment up to 25 patients per arm permitted to ensure an adequate number of evaluable patients.

Considering the possibility that the distribution of biomarker levels could be skewed, the Kolmogorov-Smirnov test of normality was applied to the data prior to analysis. For skewed biomarkers, the Wilcoxon rank sum test was used to compare expression levels between calcitriol and placebo-treated patients.

A balanced randomization procedure was used to allow both groups to be balanced with respect to preoperative Gleason score. Gleason scores were divided into four strata: 2-4, 5-6, 7, and 8-10. Treating physician and examining pathologist were blind to randomization status.

Forty-two patients were enrolled in this study between October 1999 and March 2002. Of these, 40 received treatment, 1 withdrew consent, and 1 discontinued participation for unrelated reasons. Of the 40 patients who received treatment, 2 withdrew prematurely (1 due to nausea, the other due to patient preference); 1 was not evaluable because of delay of surgery unrelated to the study. Thus, 39 were evaluable for toxicity and 37 were evaluable for molecular end points. Baseline characteristics of patients are summarized in Table 2.

Due to the variability in the volume of carcinoma in paraffin blocks, some patients' neoplastic tissue was exhausted by the application of multiple antibodies, accounting for the difference in the number of cases available for biomarker analysis.

No deaths have occurred. Toxicity is detailed in Table 3. One placebo-treated patient withdrew due to a grade 2 toxicity of vomiting. All remaining toxicity was grade 1. No significant changes in serum calcium were detected. The median and range of serum calcium pretherapy was 9.3 (8.5 to 10.4) mg/dL and was 9.3 (8.6 to 10.1) mg/dL on day 3, 9.4 (8.7 to 10.2) mg/dL on day 10, and 9.3 (7.8 to 10.0) mg/dL on day 25.

No unexpected complications from surgery were observed. Pathologic and initial clinical outcome with surgery is summarized in Table 4.

Ethanol fixation of the prostatectomy specimens was chosen because this approach has been shown to yield DNA, RNA, and protein that is superior to formalin-fixed tissue (42). Although it has been suggested that ethanol fixation does not interfere with routine immunohistochemistry (42), our experience suggests that some antigens of interest are not optimally preserved by this approach. Specifically, antibodies to Ki-67, p21, and p27 obtained from several commercial sources did not yield reliable signals on ethanol-fixed tissue, despite application of a variety of antigen-retrieval methods. Thus, our experience suggests that not all proteins can be studied immunohistochemically in ethanol-fixed tissue, and it may be necessary to analyze expression of these genes at the mRNA level instead (e.g., in situ hybridization). Biomarker results are reported in Table 5.

VDR expression was uniformly detected in the calcitriol group and the placebo group. In placebo group samples, VDR expression could be detected in essentially all adenocarcinoma cells (median, 100%; mean, 98.6%). No placebo-treated sample had less than 83.9% VDR positive adenocarcinoma cells (Fig. 1).

In samples from calcitriol-treated patients, the fraction of adenocarcinoma cells expressing VDR was reduced (median, 91.7%; mean, 75.3%). This reduction was significant (P = 0.004). The fraction of VDR positive adenocarcinoma cells in the calcitriol-treated patients' samples was also more variable than in placebo-treated patients. The difference in VDR expression between the calcitriol group and the placebo group was not due to a difference in tumor grade between the two groups. The two groups were well balanced as to both biopsy and surgical Gleason score. VDR expression in the samples was independent of Gleason score (P = 0.6).

Calcitriol treatment did not result in a statistically significant change in the fraction of cells expressing TGFβ RII, PTEN, or PCNA. The fraction of cells expressing detectable TGFβ RII and PTEN were higher in calcitriol-treated patients' samples than in placebo-treated patients' samples, as we hypothesized, but these increases did not reach statistical significance. The expression of TGFβ RII, PTEN, and PCNA were all highly variable; consequently, this study was underpowered to test whether these differences were true findings. We found little Bcl-2 or c-Myc expression in this study. Although other investigators have reported increased Bcl-2 expression in the range of >30% in prostatic adenocarcinomas (43), our finding of absent staining for Bcl-2 is similar to that reported by Haussler et al. (44) and Rakozy et al. (45), who reported 3% and 5% expression, respectively. Regarding c-Myc expression, our experience was similar to that of Cohen et al. (46), who found c-Myc only rarely expressed by immunohistochemistry. Other methodologies, such as fluorescence in situ hybridization (47) and Northern blotting (48) have reported higher yields of the proto-oncogene and suggest an association of c-Myc expression with more aggressive disease.

Prostate cancer is the most commonly diagnosed malignancy and the second leading cause of male cancer deaths in the United States. The impact of prostate cancer on public health is likely to increase as the population ages. No universally accepted prevention strategy capable of reducing the burden of this disease has been developed. The androgen signaling pathway is an attractive target for prostate cancer prevention and it has recently been reported that therapy with finasteride, an inhibitor of type II 5-α reductase, reduces the incidence of prostate cancer by approximately 25%, although the incidence of high-grade cancers were increased (49). Further development of effective chemoprevention strategies is urgently needed. Evaluation of surrogate biomarkers has been proposed as a method to select the most promising agents for chemoprevention trials. This new strategy has been used to study lycopene (50), dietary fat restriction, flax seed supplementation (51), and fenretinide (52).

One of the goals of this effort was to test the feasibility of preoperative intervention as an investigative model in prostate cancer. As illustrated by our results, preliminary data regarding individual biomarker signal level and variability is critical to the design of preoperative trials that measure biological end points. As is the case for VDR in our study, when biomarker expression is relatively uniform, observations can be made in a relatively small number of patients. In the case of biomarkers that are less frequently expressed and where expression is more variable, a larger number of patients are required. However, even for TGFβ RII, based on the observed signal and variability, a sample size of 60 per group would have 90% power to test the hypothesis that a given agent can alter TGFβ RII expression to the degree seen in this study. Although larger sample sizes are needed when biomarker expression and treatment effects are more variable (as was the case for several biomarkers reported here), the number of patients needed remains orders of magnitude smaller than that necessary of prevention studies with clinical end points. Thus, the preprostatectomy model is a potentially valuable tool for initial examination of interventions that hold promise for prostate cancer prevention and treatment. The design of such studies is hampered, however, by a paucity of published data regarding biomarker expression in human prostates. This study provides such data for several relevant biomarkers.

Our data showed a significant reduction in VDR expression in the calcitriol group when compared with the placebo group. Experiments in preclinical systems have produced conflicting information about the impact of calcitriol on VDR expression. Many in vitro experiments that examined VDR expression after short-term (≤24 hours) exposure to calcitriol suggest that VDR concentrations increase in response to calcitriol. Results have not been uniformly consistent, however, with different methodologies and different model systems yielding different observations. Reinhardt et al. (53) showed that at 16 hours, calcitriol increased VDR measured by a cytosolic binding assay but not by whole-cell binding assay in T47D breast cancer cells and suggested that increased ligand metabolism rather than loss of receptors is responsible for the apparent decrease in VDR when the whole-cell binding assay is used. Using ligand-binding assays in T47D cells, Davoodi et al. (54) showed that VDR increased 7-fold at 4 hours, decreased at 24 hours, and returned to near baseline by 72 hours. RNase protection assays showed no change in VDR mRNA. These findings suggested that the early increase in VDR may be due to that ligand-induced stabilization of the receptor (54). Similarly, Arbour et al. (55) showed that VDR protein but not RNA increased in response to short term calcitriol exposure in rat osteosarcoma cells. Furthermore, they showed that the half-life of the receptor protein was significantly longer after calcitriol treatment (55). Interestingly, VDR protein concentrations were greatest at 12 hours and declined by 24 hours. Others reported similar observations in COS-1 cells and yeast cells transiently transfected with a VDR expression vector (56), human keratinocytes (57, 58), and chondrocytes (59). In contrast to these results, Cornet et al. (60) reported a significant increase in VDR mRNA after 24 hours of calcitriol treatment in rat primary cultures of Schwann cell. Others made similar observations in HL-60 cells (61) and 3T3-L1 preadipocytes (62). Interestingly, Zhao et al. (63) reported that differentiation could alter the VDR response to calcitriol in HT-29 human colon cancer cells. VDR binding and mRNA levels increased transiently in response to calcitriol in HT-29 cells but decreased in response to calcitriol in HT-29 cells that had been treated with one of several differentiating agents (63). Similar results were reported in HL-60 leukemia cell sublines (64). Finally, in human megakaryocytic leukemia cells, calcitriol produced rapid down-regulation of VDR mRNA (65). Thus, most, but not all, in vitro studies show that calcitriol treatment increases VDR transiently and that the effect lasts ≤24 hours. There is no uniform agreement regarding the mechanism underlying these observations: some studies point to receptor stabilization, whereas others suggest a role for increased transcription. Short-term in vivo studies yield similar results. For example, VDR mRNA and protein are markedly increased in the rat intestine 6 and 12 hours after calcitriol administration but return to baseline after 24 hours (66). Longer term effects of calcitriol on VDR expression in vitro are not well characterized. Longer term treatment in animals (14 days in C57BL6 mice and Lewis rats), was associated with significant reduction in VDR protein expression (67).

None of the preclinical models replicate the treatment schedule tested in this study, in which VDR expression was tested 4 days after calcitriol therapy. Indeed, although our study shows that high-dose calcitriol down-regulates VDR in human prostate cancer, this finding is not inconsistent with brief early up-regulation seen in many preclinical models. Without sampling the tumors at an earlier time point, we cannot determine if VDR expression was first increased and subsequently down-regulated. Our results are also consistent with those seen in animal models treated with calcitriol for extended periods. The finding that high-dose calcitriol reduces VDR expression in human prostate adenocarcinoma has not been previously reported and should be confirmed in additional studies.

We cannot determine if the observed VDR down-regulation is due to differentiation as has been described in HT29 cells (63). We also cannot determine, from this investigation, the mechanism of this effect. Gene inactivation, reduced receptor stability, and functional inactivation could all play a role in such effects and should be considered in additional studies.

Overall, the reduction in the mean fraction of adenocarcinoma cells expressing VDR in the calcitriol-treated patients was modest. However, whereas nearly all of the adenocarcinoma cells in the placebo-treated samples expressed detectable VDR, the fraction of adenocarcinoma cells in the calcitriol-treated samples that expressed VDR was both reduced and highly variable. This might result from sensitivity to VDR down-regulation by calcitriol that varies from tumor to tumor, differences in the clearance of high-dose pulse calcitriol or other unknown factors. An interval of 3 to 4 days between the last calcitriol dose and surgery was chosen to allow return of calcitriol blood levels to normal prior to surgery (16). However, the time course of the down-regulation of VDR following calcitriol therapy is unknown and should be explored in future confirmatory studies.

It is important to note that VDR expression was present in all 33 prostates examined in this analysis. It is surprising, in view of the interest in the potential relationship between vitamin D and the risk of prostate cancer, that VDR expression has not previously been characterized in a large series of human prostate cancer specimens. In the normal prostate, VDR expression is lowest in men under the age of 20, increases through the sixth decade of life, and declines in the seventh decade. Further VDR expression is greater in the peripheral zone of the prostate than the central zone (68). Thus, VDR expression in the prostate seems to be greatest in the region of the prostate that commonly produces carcinomas and is higher at the time of life when prostate carcinogenesis is most active. It is not known if VDR expression in prostate cancer is greater than VDR expression in the normal prostate. However, in some tumor types including breast, cervical, and ovarian carcinomas (69), basal cell carcinomas of the human skin (70) and pancreatic cancer (71), VDR expression is higher in cancer as compared with surrounding normal tissues. The expression of 1-α-hydroxylase, the enzyme that catalyzes the final step in calcitriol synthesis, is decreased in prostate cancer primary cultures and cell lines as compared with primary cultures of normal prostatic epithelial cells (72). One might formulate a hypothesis that tumor VDR expression is up-regulated in response to decreased local production of 1,25-dihydroxyvitamin D and that exogenous administration of calcitriol reverses this process. Further work will be necessary to test this hypothesis.

In conclusion, our study shows that in human prostate tumors, high-dose calcitriol significantly down-regulates VDR expression. Further study is needed to determine the effect of VDR down-regulation on prostate cancer progression. This study also shows that the use of the preprostatectomy model is feasible and can be used to test the effect of candidate chemopreventive agents on prostate cancer. Additional studies will be needed to validate this model as a useful predictor of successful chemoprevention.

We thank Karen Shoenbrun for preparation of study drug and randomization and Scott Cruickshank for statistical advice.