We thank Christa Kramer, Valerie Grant, Kim McMonigal, and Patricia Albrecht for patient identification and data extraction; Gloria Niehans for archival tissue samples; Julia Nguyen for technical support; Badrinath R. Konety and Waddah B. Al-Refaie for critical review of the manuscript; and Michael J. Franklin for manuscript editing.
Preclinical studies show that opioids stimulate angiogenesis and tumor progression through the mu opioid receptor (MOR). Although MOR is overexpressed in several human malignancies, the effect of chronic opioid requirement on cancer progression or survival has not been examined in humans.
We performed a retrospective analysis on 113 patients identified in the Minneapolis VA Tumor Registry (test cohort) and 480 patients from the national VA Central Cancer Registry (validation cohort) who had been diagnosed with stage IV prostate cancer between 1995 and 2010 to examine whether MOR expression or opioid requirement is associated with disease progression and survival. All opioids were converted to oral morphine equivalents for comparison. Laser scanning confocal microscopy was used to analyze MOR immunoreactivity in prostate cancer biopsies. The effects of variables on outcomes were analyzed in univariable and multivariable models.
In patients with metastatic prostate cancer, MOR expression and opioid requirement were independently associated with inferior progression-free survival (hazard ratio [HR] 1.65, 95% confidence interval [CI] 1.33-2.07, P<.001 and HR 1.08, 95% CI 1.03-1.13, P<.001, respectively) and overall survival (HR 1.55, 95% CI 1.20-1.99, P<.001 and HR 1.05, 95% CI 1.00-1.10, P = .031, respectively). The validation cohort confirmed that increasing opioid requirement was associated with worse overall survival (HR 1.005, 95% CI 1.002-1.008, P = .001).
Opioid requirement is strongly associated with pain levels in patients with cancer. Severe cancer-related pain requires treatment with escalating doses of opioids. Preclinical studies from our laboratory and others show that morphine at clinically relevant doses stimulates angiogenesis and promotes tumor growth in mice.[3-5] Mechanistically, morphine stimulates the growth-promoting mitogen-activated protein kinase/extracellular signal regulated kinase pathway and the survival-promoting Akt/protein kinase B pathway in endothelial and tumor cells.[4-7] Additionally, morphine activates cycloxygenase-2 (COX-2) and transactivates receptor tyrosine kinases[3, 6] in a variety of cells associated with cancer progression, including tumor and endothelial cells and pericytes.[8-11] Several studies have shown that morphine activates these signaling pathways via the mu opioid receptor (MOR).[5, 6, 12] Notably, MOR is also the receptor that mediates the analgesic activity of morphine and its congeners.
MOR is expressed not only in the nervous system but also in peripheral tissues including human lung cancer and colon cancer.[5, 6, 12, 13] We and others have observed that MOR immunoreactivity (MOR-ir) is higher in human lung cancer biopsies than in healthy lung tissue.[5, 6] In patients, positron emission tomography scans showed that binding of the MOR agonist [11C]carfentanil was higher in lung cancer tissue than in healthy lung tissue. Polymorphisms in the MOR gene appear to be clinically relevant, since breast cancer patients with 1 or more copies of the G allele at A118G in the MOR gene have decreased receptor transcription, diminished response to opioid binding, and better survival than those with the A/A genotype, and the G/G genotype is associated with a lower risk of developing esophageal cancer.
Because of MOR expression in tumors, it is possible that endogenous ligands and/or opioids used for analgesia may inadvertently promote cancer progression. This hypothesis is supported by the following observations: 1) pain is an independent prognostic factor for overall survival (OS) in humans with castrate-resistant prostate cancer or non–small cell lung cancer[1, 16-19]; 2) the MOR antagonist naltrexone improved survival in patients with malignant astrocytomas treated with radiation; 3) OS was superior in a subset of patients with advanced pancreatic cancer who received surgical celiac plexus block with alcohol (versus saline control) and required fewer systemic opioids; and 4) a randomized clinical trial in 202 patients with various advanced solid tumors experiencing refractory pain showed that an implantable intrathecal opioid delivery device reduced the need for systemic opioids and was associated with a trend toward improved OS (P = .06).
It is not known whether long-term opioid requirement independently influences cancer progression, metastasis, or OS in patients. We performed a retrospective analysis to examine the hypothesis that the dose of opioids required and/or MOR expression is associated with cancer progression and survival in metastatic prostate cancer. We selected prostate cancer as a model because it is one of the most commonly diagnosed cancers and because patients with advanced disease frequently experience painful bone metastases and require long-term opioid treatment.
PATIENTS AND METHODS
We analyzed the data for 113 patients who were diagnosed with stage IV prostate cancer between 1995 and 2008 at the Minneapolis Veterans Affairs (VA) Health Care System (MVAHCS) and 480 patients from the national VA Central Cancer Registry (VACCR) diagnosed between 2004 and 2010, all of whom were treated with first-line androgen deprivation therapy, to determine whether baseline MOR expression or ongoing opioid requirement for cancer-related pain was associated with inferior outcomes. Demographic, clinical, and pharmacological data were obtained from patient records, the tumor registry, and VA Data Support Services. The study was approved by the institutional Human Subjects Committee.
Test and Validation Cohorts
The test cohort consisted of 113 patients from the MVAHCS. Progression dates were calculated using manual prostate-specific antigen (PSA) review and chart abstraction. Progression was defined as PSA progression (based on Prostate Cancer Clinical Trials Working Group guidelines) or need for palliative radiation, whichever occurred first. Data on the 480 patients in the validation cohort were obtained from the national VACCR. All patients from the MVAHCS were excluded from the validation cohort.
All oral and transdermal outpatient opioid prescriptions dispensed from any VA in the United States between 1995 and 2012 were collected to determine the total opioid quantity dispensed per prescription. All opioid prescriptions were converted to oral morphine equivalents (OME) using an equi-analgesic conversion table. The average opioid requirement was calculated for 2 distinct treatment intervals: 1) 12 months prior to diagnosis (opioid requirement before diagnosis) and 2) from diagnosis to death/last follow up (opioid requirement since diagnosis).
Laser Scanning Confocal Microscopy for MOR-ir
MOR-ir was determined on 6-μm sections of archived, paraffin-embedded tumor specimens from 61 of the 113 patients in the MVAHCS cohort. Samples of benign prostatic hyperplasia (BPH) used as controls were obtained from patients without prostate cancer. Briefly, after antigen retrieval with 10 mM sodium citrate (pH 6.0) at 96°C for 30 minutes, sections were immunostained with 1:100 guinea pig anti-human MOR (Millipore, Billerica, MA) and 2° donkey anti-guinea pig Cy5 (1:200, Jackson Immunoresearch, West Grove, PA). In parallel, sections were stained with isotype-matched immunoglobulin G and 2° antibody for negative controls. Z-stack images (0.5 μm thick) were acquired using an Olympus Fluoview FV1000 BX2 Upright laser scanning confocal microscope BX61 (Olympus Corporation, Center Valley, PA) with a 60×/1.42 oil objective lens. Images were digitized through the Fluoview FV1000 view program, and fluorescence acquired in the Cy5 channel was pseudocolored red to show MOR staining. Images were binarized and analyzed with Adobe Photoshop and Image Processing Tool-kit Plug-in Functions (Reindeer Games, Asheville, NC) to quantitate MOR-ir as a marker of MOR expression. Data are presented as the percentage of positive pixels of MOR-ir in the total stained area. Patients were stratified into low MOR (less than the median value of 2.19% MOR-ir positive pixels) and high MOR (greater than or equal to the median value of 2.19% MOR-ir positive pixels) groups.
A multivariable Cox proportional hazards regression model was used for progression-free survival (PFS) and OS, whereas a competing risks model was used for time to progression (TTP). Covariates included opioid requirement prior to diagnosis, opioid requirement since diagnosis, age, PSA at diagnosis, Gleason score, year of diagnosis, and number of different metastatic sites (eg, bone, lung, liver) at diagnosis for the MVAHCS (test) cohort. Analysis of the national (validation) cohort did not include PSA, as this data was not available from the national VACCR. A MOR-ir variable was added for the subset of patients (n = 61) in the test cohort for whom these data were available. For these models, opioid requirement was taken in increments of 5 mg/d OME, and PSA was taken in increments of 10 ng/mL. All analyses were performed using SAS version 9.2 (SAS Institute, Inc., Cary, NC) and R version 2.9.2 (R Development Core Team, 2009, Vienna, Austria). P<.05 was considered significant.
MOR-ir in the Test Cohort
To determine whether MOR-ir was associated with outcomes in patients with newly diagnosed metastatic prostate cancer who were receiving first-line androgen deprivation therapy, we assayed tumor tissues for MOR-ir. The median MOR-ir pixel ratio in prostate cancer samples was 2.19 (range, 0.95-6.82), whereas the median MOR-ir pixel ratio in BPH samples (controls) was 0.56 (range, 0.16-1.00; Fig. 1A). Intriguingly, in the high MOR samples, almost every cell showed bright MOR-ir that colocalized with the cell membrane. However, in the low MOR biopsies, MOR-ir was randomly scattered and often observed in the nuclei (Fig. 1A, enlarged inset). BPH samples showed MOR-ir in fewer cells than in prostate cancer tissue from the low MOR group. It is possible that colocalization with the membrane may be indicative of an active receptor, whereas nuclear MOR-ir may be an inactive internalized receptor.
Univariable Analysis of MOR-ir in the Test Cohort
On univariable analysis, patients with low MOR-ir had markedly superior TTP, PFS, and OS compared with patients who had high MOR-ir (Fig. 1B). The median values for TTP, PFS, and OS were all greater in the low MOR-ir group compared with the high MOR-ir group (1.80 vs 0.53 years, Gray P<.001; 3.31 vs 1.07 years, Gray P<.0001; and 3.82 vs 1.14 years, Gray P<.0001, respectively). Thus, high MOR-ir levels were strongly associated with worse clinical outcome.
Multivariable Analysis of Opioid Requirement in the Test Cohort
Patients in the metastatic cohort used a wide variety of opioid medications, most commonly oxycodone (53%), hydrocodone (50%), codeine (49%), and morphine (30%). Hydromorphone, fentanyl, and methadone were used infrequently. Long-term use of opioids prior to diagnosis was infrequent (only 3 patients used >5 mg/d OME). In the multivariable model (Table 1), opioid requirement after diagnosis (as a continuous variable in increments of 5 mg/d OME) was associated with shorter PFS (HR 1.08, 95% CI 1.04-1.11, P<.001) and worse OS (HR 1.07, 95% CI 1.04-1.11, P<.001). Opioid requirement was not associated with TTP; however, 25 of 113 (22%) patients died before evidence of PSA progression, and 9 of 113 (8%) patients received palliative radiation in the first 90 days following diagnosis, which likely affected subsequent PSA trends that were used to determine TTP. The year of diagnosis did not influence any of the outcomes; furthermore, the impact of race could not be evaluated, because only 4% of our prostate cancer patients identified themselves as African American. Therefore, neither of these factors was included in the final multivariable model. Our results suggest that opioid requirement is independently associated with worse PFS and OS in patients with metastatic prostate cancer.
Table 1. Multivariable Analysis for Time to Progression, Progression-Free Survival, and Overall Survival in Patients With Hormone-Sensitive Metastatic Prostate Cancer in the Test Cohort
Multivariable Analysis of MOR-ir and Opioid Requirement in the Test Cohort
Because MOR-ir was strongly associated with clinical outcomes on univariable analysis, we included this factor in the multivariable model in the test cohort, together with opioid requirement and known prognostic factors (Table 1). Taken as a continuous variable, higher MOR-ir remained significant for PFS (HR 1.65, 95% CI 1.33-2.07, P<.001) and OS (HR 1.55, 95% CI 1.20-1.99, P<.001) but not for TTP. For every one unit increase in the MOR-ir positive area (eg, from 3% to 4% positive pixels in the total stained area), the risk of progression/death increased by 65%, and the risk of death increased by 55%. Additionally, for every 5 mg/d increase in opioid requirement after diagnosis, the risk of progression/death increased by 8% (HR 1.08, 95% CI 1.03-1.13, P<.001) and the risk of death increased by 5% (HR 1.05, 95% CI 1.00-1.10, P = .031). We found no evidence that the effect of MOR-ir on the outcome depended on opioid requirement, so an interaction term was not used in the final model. Taken together, these results suggest that both MOR-ir and opioid requirement are independently associated with worse PFS and OS in patients with metastatic prostate cancer.
Multivariable Analysis of Opioid Requirement in the Validation Cohort
In the validation cohort, outcome data was available for OS but not for TTP or PFS, and pathological material was not available for assessment of MOR-ir. As observed in the test cohort, opioid requirement after diagnosis was significantly associated with shorter OS (HR 1.005, 95% CI 1.002-1.008, P = .001) in the independent validation cohort (Table 2). Advanced age and higher Gleason score were also independently associated with worse OS, further supporting the results observed in the test cohort.
Table 2. Multivariable Analysis for Overall Survival in Patients With Hormone-Sensitive Metastatic Prostate Cancer in the Validation Cohort
This is the first report of an association between MOR expression, long-term opioid requirement, and cancer outcomes. Greater opioid requirement and increased MOR-ir in tumor tissue were associated with worse PFS and OS in patients with metastatic, hormone-sensitive prostate cancer. Interestingly, when both MOR-ir and opioid requirement were included in the multivariable model, the significance of several other known prognostic variables (number of metastatic sites, Gleason score, and age) diminished considerably.
In our study, MOR colocalized with the cell membrane in prostate cancer biopsies with strong MOR-ir, but it colocalized with the nucleus in prostate cancer biopsies with weak MOR-ir. This finding is consistent with other reports in colon and lung cancer. Resected human colonic tumor and nontumor biopsies from a region 10 cm away from the tumor showed membrane and nuclear MOR expression, respectively. This previous study found that morphine augmented the secretion of urokinase type plasminogen activator in the human colon cancer cell line HT-29, suggesting that morphine may promote metastases. It is likely that membrane-associated MOR is active and stimulates downstream signaling, whereas nuclear MOR is the internalized, inactive receptor that has neither been degraded nor recycled. It is possible that many tumors express MOR, which may result in increased MOR-induced signaling upon opioid exposure.
Experimental studies provide insight in the mechanisms by which MOR may contribute to cancer progression and metastases. MOR overexpression in human bronchoalveolar lung carcinoma cells led to increased tumor growth and lung metastases in nude mice compared with vector-transfected cells. Blocking MOR with methylnaltrexone, MOR silencing on tumor cells, or MOR deletion in mice led to reduced lung tumor growth and metastases. Because MOR overexpression on tumor cells promoted tumor growth in nude mice, the effect of this pathway was not dependent on host immune function. However, endogenous mu opioid peptides secreted by other malignant cells modulate immune response via MOR. MOR-mediated disruption of endothelial barrier function[11, 26] may also possibly contribute to the dissemination of malignant cells to the bone marrow, as found in early stage prostate cancer. Indeed, MOR antagonists abrogated metastases and increased survival in a murine neuroblastoma model. Furthermore, naltrexone markedly reduced human squamous cell carcinoma xenograft growth in nude mice. Our clinical observation that higher MOR expression correlates with poorer outcomes in prostate cancer is consistent with these experimental studies, suggesting that MOR contributes to cancer progression. The fact that the MOR antagonist naltrexone induces clinical response in prostate cancer30 further supports this notion. These observations suggest that selective peripheral MOR antagonists administered alone (or coadministered with opioids) may prevent the adverse effects of MOR activation on tumors, without compromising analgesia.
Although the present study suggests that high opioid requirement is associated with cancer progression, other studies suggest that pain may promote cancer growth via cyclo-oxygenase–mediated prostaglandin release.[31, 32] Recent studies have shown that baseline pain is a prognostic factor for survival in patients with castration-resistant prostate cancer and other solid tumors, but most of these studies did not include an analysis of opioid requirement.[1, 16, 19, 33, 34] We have demonstrated that morphine-induced breast tumor growth in mice was accompanied by increased pain (over time), increased metastases, and reduced survival. COX-2 expression was elevated in endothelial and nonendothelial cells in the tumors of morphine-treated mice, and the COX-2 inhibitor celecoxib inhibited morphine-induced tumor growth and metastases and increased survival without compromising analgesia. Morphine-induced COX-2/PGE2 expression in the spinal cord has been implicated in opioid analgesia. There are few prospective clinical data examining the effect of ongoing (chronic, rather than just at diagnosis) cancer pain on clinical outcome. The lack of availability of pain scores in our dataset precluded inclusion of this factor in the present study. Because high opioid requirement correlated with reduced PFS and OS, it is possible that this group had more pain and activation of COX-2 and other signaling pathways, contributing to poor disease outcome and increased opioid requirement.
It is important to distinguish chronic opioid use in patients with advanced cancer (as examined in the current study) from short-term opioid use in the perioperative period, and from chronic opioid use in patients without malignancies. The pharmacological and biological effects of opioids and their receptors, as well as immune function, are likely different in these various situations, making extrapolation from one scenario to the other difficult.
Some,[36-38] but not all, retrospective studies examining the effect of anesthetic techniques on outcome following surgery for prostate, breast, or colon cancer reported that recurrence/survival was worse in patients receiving general anesthesia with short-term systemic opioid analgesia compared with patients receiving epidural anesthesia/analgesia. The only prospective randomized trial reported found that recurrence-free survival was not different in patients receiving epidural versus systemic anesthesia/analgesia for abdominal surgery for various malignancies. Although outcomes were better in some retrospective studies in which patients in the epidural group received less systemic opioids, it is challenging to distinguish the independent effects of opioids from the effects of anesthetic techniques. The nature, cause, and mechanism of any potential effect of anesthetic technique and perioperative analgesia on cancer outcomes require further investigation.
It is uncertain whether chronic opioid use prior to a diagnosis of cancer influences the behavior of the tumor. We have observed that long-term opioid requirement for non–cancer-related pain (e.g., osteoarthritis) prior to diagnosis did not affect the stage, grade, or treatment outcomes in patients diagnosed with early stage prostate cancer (unpublished observations). However, a recent study reported that prostate cancers diagnosed in opium addicts had higher Gleason scores than those in nonaddicts. These findings indicate a need for larger, prospective studies examining the association of MOR expression and opioids with cancer development, behavior, and outcomes.
The above observations are consistent with the notion that the tumor microenvironment in established cancers may induce increased responsiveness to opioids, likely via increased MOR expression. The tumor microenvironment is replete with proinflammatory and growth-promoting cytokines that regulate MOR expression. We observed previously that vascular endothelial growth factor induced MOR expression on mouse retinal microvascular endothelial cells. We also found that the human non–small cell lung cancer cell line H-2009 secreted significantly higher levels of cytokines as compared to control Beas2B lung epithelial cells. Moreover, H2009 conditioned media, containing secreted cytokines, stimulated MOR expression on Beas2B cells. Additionally, we observed a several-fold higher secretion of the cytokines vascular endothelial growth factor and interleukin-6 from the human prostate cancer cell line DU145 compared with the benign prostatic hyperplasia BPH1 cell line (unpublished observations). Thus, it is possible that the cytokine-rich microenvironment in established prostate tumors stimulates the observed increase in MOR-ir and thereby enables the tumor to respond to opioids. Opioid responsiveness may not be so evident in patients receiving opioids prior to tumor formation, or after potentially curative resection or radiation treatment.
In conclusion, our study demonstrates an association between MOR expression, long-term opioid requirement, and prostate cancer outcomes in humans. We found that patients with metastatic, hormone-sensitive prostate cancer with high baseline MOR expression or patients who require higher doses of opioids have significantly shorter PFS and OS. These findings indicate that MOR expression may be a predictor of long-term survival in prostate cancer, independent of other key factors, and could be used to guide treatment decisions if validated in larger studies. The limitations of the current study include: 1) its retrospective nature; 2) the use of the quantities of opioids dispensed as a surrogate measure of actual opioid ingestion; 3) the potential confounding effect of tumor volume/burden on pain, opioid use, and outcomes; and 4) the restriction of the analysis to a particular situation (i.e., patients with metastatic prostate cancer treated with first-line androgen deprivation therapy). Therefore, further investigation is required before reaching conclusions that alter pain management strategies in cancer patients. Prospective, randomized trials of opioid-sparing strategies (such as coadministration of peripheral opioid antagonists or the use of nonopioid analgesics) to treat cancer-related pain will be required to determine whether there is a causal relationship between opioids and PFS or OS, and whether such strategies will improve clinical outcomes. Clinical practice should not be changed until the results of such trials are available.
This study was supported by the Veterans Health Administration (P.G.) and by National Institutes of Health grants R01 CA109582, R01 HL68802, and R01 HL103733 (K.G.); R01 HL68802-06S1 (D.V.); and T32 HL007062-34 (D.Z.).