The study was comprised of 1,034 study subjects (prostate cancer cases = 499, controls = 535). Patient demographics and serum PSA of cases and controls are shown in . Most patients were Caucasian (81.0%) and over 60 years of age (73.3%). The distribution of serum PSA in the cases reflects our selection of patients meeting PSA criteria for advanced/aggressive disease (PSA > 10 ng/mL).
Characteristics of study population (advanced /aggressive prostate cancer cases and controls).
shows the demographics and disease characteristics of cases, comparing NSAID-users to non-users. Compared to those patients who did not use NSAIDs, NSAID users were more often Caucasian, were better educated, and were heavier. Although the distribution of clinical stage and Gleason score was relatively similar between the two groups, there was a non-significant trend towards lower serum PSA in NSAID users. These findings suggest that NSAIDs may act to suppress serum PSA, a phenomenon that has been previously reported by several different groups [35
]. The decision to perform prostate biopsy is typically triggered by PSA elevation above a certain threshold. Thus, a PSA-suppressing effect of NSAIDs could potentially delay biopsy and subsequent prostate cancer diagnosis in NSAID users. As the clinical ramifications of delayed diagnosis are significant, the potential for PSA suppression by NSAIDs certainly deserves further investigation.
Characteristics of Aggressive Prostate Cancer Cases (NSAID users compared to non-users).
Of the 1,034 subjects, 612 (59.2%) reported a history of NSAID use. Among these men, ASA use was more common than ibuprofen use (88.2% vs.
24.2%); some patients used both ASA and ibuprofen (12.4%). The association between NSAID use and the risk of advanced/aggressive prostate cancer is shown in . As noted previous, a decreased risk of disease was seen in both NSAID users (OR 0.67, 95% CI 0.52–0.86) and ASA users (OR 0.66, 95% CI 0.51–0.86), relative to NSAID non-users [15
]. This decrease was also seen across increasing quintiles of intake. There was a trend towards a decreased risk of prostate cancer in ibuprofen users, although this did not reach statistical significance (OR 0.83, 95% CI, 0.56–1.21). This lack of significance may be due to the relatively small number of ibuprofen users in the study.
Odds ratios and 95% confidence intervals for aggressive prostate cancer by use of NSAID medications.
The dose and duration of NSAID use was also found to impact the risk of advanced/aggressive prostate cancer. shows the risk of disease in NSAID and ASA users divided into quintiles based on extent of NSAID use. A similar analysis of ibuprofen use was not possible due to the small number of ibuprofen users. In both NSAID and ASA users, a stepwise decrease in the risk of prostate cancer was seen with increasing dose and duration of medication use. Compared to the first quintile, a significantly decreased risk of disease was seen in subjects in the fifth for both NSAID (OR 0.58, 95% CI 0.39–0.85) and ASA (OR 0.60, 95% CI 0.40–0.90).
Odds ratios and 95% Confidence Intervals for aggressive prostate cancer by pills per year of NSAID use.
We have previously reported an inverse association between NSAID use and risk of advanced/aggressive prostate cancer [10
]. A similar, protective effect of NSAIDs has been reported by others [11
], but remains controversial [22
]. However, the majority of these studies investigated the impact of NSAIDs on all prostate cancer diagnoses [11
] and few studied the association with aggressive disease [13
]. Of the studies specifically investigating advanced or aggressive disease, most reported a protective effect of NSAIDs. Norrish et al
. identified a stronger inverse association in “advanced” cases, defined as extracapsular disease extension or Gleason score ≥ 7, relative to all prostate cancer cases [18
]. Leitzmann et al
. reported no association between NSAID use and all cancers, but a trend towards a decreased risk of metastatic disease was noted in ASA users [25
]. Mahmud et al
. found that regular NSAID use was inversely related to the risk of detection of poorly differentiated cancers or those with higher percentage of core involvement on prostate biopsy [17
]. The current study, restricted only to men with advanced/aggressive disease, shows a strong protective effect of NSAIDs on disease risk. These findings are interesting, as they suggest NSAIDs may preferentially inhibit the development of high-grade tumors, or prevent progression to advanced stage disease.
The most intriguing results of our genetic analysis are shown in . For each inflammatory gene polymorphism, we first present their main association with prostate cancer. Next, patients were stratified by genotype, and the risk of disease in NSAID users relative to non-users was calculated. Genetic variants in several inflammatory pathway genes appeared to modify the protective effect of NSAIDs on prostate cancer development. For several SNPs, NSAID use reduced the risk of prostate cancer only in carriers of a particular genotype, whereas no association was seen in carriers of alternative genotypes. For example, in carriers of the AC + AA MSR1 rs10503574 genotype, NSAID use markedly decreased the risk of prostate cancer (OR 0.31, 95% CI 0.15–0.62), whereas NSAIDs did not appear to alter disease risk in carriers of the AA genotype (OR 0.79, 95% CI 0.60–1.04) (p-interaction < 0.01). For other SNPs, a protective effect of NSAIDs was seen regardless of genotype, however, the specific genetic variants modified the magnitude of this effect. The remainder of the SNPs did not appear to materially alter the impact of NSAIDs on disease risk (not shown; available upon request).
These data suggest that NSAID use may not result in a uniform protective effect across all individuals. The variation of NSAID effect according to genotype may further explain the conflicting results of previously published studies investigating NSAIDs and prostate cancer risk. Depending on the genetic makeup of study participants, NSAIDs may have a largely protective effect among certain individuals, but have no influence on prostate cancer development in others. This could impact any future recommendations that regular NSAID be used for prostate cancer prevention, as one might need to know an individual’s genotype before making such a recommendation.
The focus of this paper is on the potential effect modification of our previously reported inverse association between NSAID use and aggressive prostate cancer by genes that were marginally associated with disease. A number of statistical comparisons were made and fifty-two out of 311 SNPs exhibited associations with aggressive prostate cancer; of these 52, six (12%) showed potential interactions with NSAID use on risk of prostate cancer (). The number of associations observed here are substantially more than that expected based on chance alone. Since our aim was to generate hypotheses we have presented the nominal p-values. Nevertheless, in light of the multiple comparisons undertaken here these require further confirmatory work for confirmation
While many have speculated that chronic inflammation plays a key role in prostate carcinogenesis [39
], the molecular mechanisms through which this effect occurs remain unclear. The genetic enzymes reported here may act as the molecular targets for NSAIDs and their metabolites, and further investigation of these proteins may shed light on the mechanisms through which NSAIDs may impact prostate cancer.
A number of these genes are well characterized and have been previously linked to prostate carcinogenesis. Increased COX-2
expression has been reported in prostate tumors [40
], and may be associated with high-grade lesions[45
] or prostate cancer progression [46
]. Others have reported that COX-2
expression is not consistently elevated in prostate cancer cells themselves, but instead is overexpressed only in inflammatory cells or atrophic epithelial cells within the prostate [34
]. Nonetheless, genetic variation in COX-2
has been shown to alter the risk of prostate cancer irrespective of NSAID use. TLR4
is a receptor molecule that initiates pro-inflammatory signaling pathways upon recognition of pathogen associated molecular patterns. Prior studies have reported differential TLR4
expression in high-grade tumors relative to normal prostate tissue [47
], and sequence variants in several TLR
genes have been linked to prostate cancer risk [48
β is thought to have an anti-inflammatory function, and it appears that genetic variation of this enzyme alters prostate cancer risk [49
, an activator of natural killer cells and T-cells, has shown anti-tumor effects in a number of different cancer models [50
]. In mice, IL-12
transduced bone marrow cells were shown to suppress the development of metastatic prostate cancer [53
]. Mutations of MSR-1
, a transmembrane protein expressed on macrophages and involved in innate immunity, are associated with both hereditary and sporadic prostate cancer risk [54
Furthermore, studies of various disease processes have suggested that NSAIDs interact with several of these gene products on the molecular level. The inhibition of COX-2 by NSAIDs is well characterized, and many have hypothesized that this inhibition is responsible for the potential chemoprotective effects of NSAIDs on prostate cancer development [55
]. The gastroenterology literature has reported that NSAID-induced bowel injury may be due to an NSAID-triggered inflammatory response mediated by a TLR4-dependent signaling pathway [57
]. In models of esophageal cancer, it has been proposed that TGFβ may serve to induce COX-2 expression [58
]. Finally, experimental data has shown that NSAIDs are capable of inhibiting IL-12 folding and secretion;[59
] and NSAID treatment results in decreased IL-12 expression in cervical cancer tissue [60
]. Similar molecular interactions in the prostate may serve as the mechanism through which NSAIDs alter prostate cancer risk.
does not appear to be overexpressed in human prostate cancer cells, the protective effects of NSAID may function through COX-2
inhibition of endogenous inflammatory infiltrates, or even through a COX-2
independent pathway [34
]. Of the other five genes found to modify the effect of NSAIDs in the current study, all are expressed on monocytes, macrophages, or other inflammatory cells. Thus, the mechanism through which NSAIDs alter prostate cancer risk may be through interaction with inflammatory infiltrates within prostate tissues, and not interactions with prostate epithelial cells themselves [34
A final consideration is that inter-individual differences in NSAID metabolism may account for the differential effects of NSAIDs on prostate cancer development. Genetic variation could potentially result in differential NSAID metabolism resulting in increased intra-prostatic penetration of NSAIDs and their metabolites in certain individuals relative to others. These hypotheses warrant further research to clarify the mechanism through which NSAIDs may alter prostate carcinogenesis.
A limitation of the current study is that controls might not have been fully representative of the cases' source population because they were selected from men who underwent standard annual medical examinations. In comparison with the cases' source population, such controls might have cared more about their health, been more highly educated, had a higher income, been more likely to have medical insurance, or had medical conditions that increased their likelihood of NSAID use. As a proxy for these potential case-control differences, we adjusted our results for education and income; doing so did not materially alter our findings (not shown). Of course, since these factors are only proxies, adjusting for them may not have fully accounted for the potential case-control differences, especially with regard to medical conditions among controls. A related concern is potential bias arising from the lack of information about NSAID use other than aspirin and ibuprofen, and ensuing misclassification. If cases and controls used such medications in a similar manner, this bias would be nondifferential among cases and controls leading to an underestimation of the effect estimates. We would also expect a bias toward the null if controls were more likely to use non-measured NSAIDs than cases. Finally, there is potential for recall bias regarding the measured NSAIDs. The lack of general knowledge about the potentially protective effects of NSAIDs on prostate cancer should have made any such bias nondifferential among cases and controls. Moreover, we did not ask participants to specify the exact pill dose, which may also have resulted in nondifferential misclassification bias, and most likely an underestimate of the effect estimates.