Prostaglandin E2 (PGE2) promotes colorectal carcinogenesis. Overall, systemic PGE2 production can be assessed by measuring its major metabolite, PGE-M, in urine. We examined the potential role of PGE-M as a biomarker for colorectal adenoma risk and chemopreventive response to anti-inflammatory drugs. We conducted a prospective case–control study nested within the Nurses' Health Study. Among women who previously provided a urine sample, we identified 420 cases diagnosed with colorectal adenoma during follow-up and matched them to 420 endoscopy-negative controls. We measured urinary PGE-M using an LC/MS assay. Compared with women in the lowest quartile of urinary PGE-M, women in the highest quartile had a multivariate OR of 1.40 (95% confidence interval (CI), 0.92–2.14) for any adenoma; 0.91 (95% CI, 0.48–1.72) for low-risk adenoma (solitary adenoma <1 cm in greatest diameter with tubular/unspecified histology); and 1.66 (95% CI, 1.04–2.67) for high-risk adenoma (adenoma ≥1 cm in greatest diameter and/or tubulovillous, villous or high-grade dysplasia histology or multiple adenomas of any size or histology). Regular use of anti-inflammatory drugs (≥2 standard tablets of aspirin/NSAIDs per week) was associated with a significant reduction in adenoma risk (multivariate OR, 0.61; 95% CI, 0.43–0.87) in women with high baseline PGE-M (quartiles 2–4), but not low PGE-M (quartile 1).Urinary PGE-M is associated with an increased risk of high-risk adenoma. Anti-inflammatory drugs seem to reduce adenoma risk among women with high, but not low PGE-M. Urinary PGE-M may serve as a biomarker to define subsets of the population who may obtain differential chemopreventive benefit from anti-inflammatory drugs. Cancer Prev Res; 7(7); 758–65. ©2014 AACR.
Considerable evidence supports a causative role for inflammation in the development of colorectal neoplasia (1, 2). A principal mechanism which mediates this association is the proinflammatory enzyme, prostaglandin-endoperoxide synthase 2 (PTGS2), also known as COX-2 (3). PTGS2 is overexpressed in colorectal cancer and adenomatous polyps, the precursor for the vast majority of colorectal cancers (4). Previous randomized controlled trials have demonstrated that anti-inflammatory drugs, such as aspirin and COX-2 selective inhibitors reduce the risk of colonic adenoma (5). Moreover, in two large cohort studies, the Nurses' Health Study (NHS) and Health Professionals Follow-Up study, we have previously shown that aspirin seems to be associated with a lower risk of PTGS2-positive colorectal cancers but not PTGS2-negative colorectal cancers (6).
PTGS2 is the rate-limiting enzyme for the metabolic conversion of arachidonic acid to prostaglandins and related eicosanoids, including prostaglandin E2 (PGE2), which is thought to be the primary contributor to neoplasia development. PGE2, through several mechanisms, increases proliferation, migration, and invasiveness of cells, promotes angiogenesis, induces resistance to apoptosis, and modulates cellular and humoral immunity (7, 8).
Measurement of excreted urinary metabolites is an accurate measure of systemic prostaglandin synthesis (9). Overall, systemic PGE2 production can be estimated by measuring its major metabolite, PGE-M (11 alpha-hydroxy,9,15-dioxo-2,3,4,5-tetranor-prostane-1,20-dioic acid), in the urine.
Previous studies have shown that prediagnostic levels of PGE-M are associated with subsequent risk of colorectal cancer (10–12). Similarly, in a cross-sectional study, PGE-M levels were associated with risk of advanced and multiple adenomas (13). We conducted this study to determine if prediagnostic levels of urinary PGE-M can predict risk of adenoma within a cohort of women enrolled in the NHS who provided urine specimens before undergoing lower endoscopy. In addition, we have previously shown in this cohort that aspirin use is associated with a lower risk of colorectal adenoma (14) Thus, we also specifically examined if urinary PGE-M may predict responsiveness to aspirin chemoprevention.
Materials and Methods
The participants were drawn from an ongoing prospective cohort, the NHS, which began in 1976 when 121,700 married U.S.-registered female nurses who were ages 30 to 55 years returned an initial questionnaire that determined a variety of health-related exposures. Since 1980 and biennially thereafter, participants have provided detailed information on their medication use, including aspirin. In addition, women have reported their utilization of endoscopy and any diagnoses of colorectal polyps and colorectal cancer. Among women who reported polyps, we requested medical records, including endoscopy and pathology reports, which were subsequently reviewed by a study physician. As a percentage of total follow-up time, follow-up in the cohort has exceeded 92%.
Between 1989 and 1990, 32,826 women from the NHS returned a blood specimen on ice packs by overnight courier. From among this cohort of women, we also requested urine specimens between 2000 and 2001. We received 18,743 samples on ice packs by overnight courier. These samples were processed, aliquoted, and archived at −130°C. The baseline characteristics of women who provided urine were similar to women who did not.
For this analysis, we included women who provided a urine sample, baseline data on aspirin use, had no prior history of adenoma or cancer (except nonmelanoma skin) and subsequently underwent a lower endoscopy through June 1, 2008. Using risk set sampling, we matched case women diagnosed with colorectal adenoma through 2008 to control women who were endoscopy-negative and were not diagnosed with adenoma. We matched cases and controls on age (within 1 year), date of urine collection (within 3 months), year of endoscopy (within 2 years), and reason for endoscopy. Cases in whom adenoma were localized only to the distal colorectum were matched to controls who had either a negative sigmoidoscopy or colonoscopy. Cases in whom at least one adenoma was localized in the proximal colon were matched only to controls who had a negative colonoscopy. Eighteen matched pairs were excluded from analysis due to failed laboratory assays. This resulted in 420 cases and 420 controls for inclusion in the final analysis.
This study was approved by the Human Subjects Committee of Harvard School of Public Health and the committee on the Use of Human Subjects in Research at Partners Healthcare.
In collaboration with the Eicosanoid Core Laboratory at Vanderbilt University, archived urine specimens collected from the NHS participants in 2000 were assayed by personnel blinded to case–control status for 11 alpha -hydroxy,9,15-dioxo-2,3,4,5-tetranor-prostane-1,20-dioic acid (PGE-M) using LC/MS with slight modifications to the method previously described (15, 16). Briefly, 1 mL urine was converted to the O-methyloxime derivative and purified by C18 solid phase extraction before analysis by LC/MS. LC was performed on a 2.0 × 50 mm, 1.7 μm particle Acquity BEH C18 column (Waters Corporation). Mobile phase A was 95:4.9:0.1 (v/v/v) 5 mmol/L ammonium acetate: acetonitrile: acetic acid, and mobile phase B was 10.0:89.9:0.1 (v/v/v) 5 mmol/L ammonium acetate: acetonitrile: acetic acid. Samples were separated by a gradient of 85% to 76% of mobile phase A over 6 minutes at a flow rate of 200 μL per minute before delivery to a ThermoFinnigan TSQ Quantum Vantage triple quadrupole mass spectrometer. Urine samples from the same case–control pair were analyzed in the same batch to eliminate between-assay variability. Quality control samples were included among case–control samples. On the basis of blinded duplicate samples, the coefficient of variation was 14%. Levels of urinary creatinine were also measured to standardize values of PGE-M, which are presented as PGE-M (ng)/ creatinine (mg).
We compared means (SD) and medians (interquartile ranges) of continuous variables for case and control participants using paired t test and Wilcoxon signed rank test, respectively. We used χ2 tests to compare categorical variables. We classified cases in two groups: “high-risk” cases were defined as adenoma ≥1 cm in greatest diameter and/or tubulovillous, villous or high-grade dysplasia histology or multiple (≥2) adenomas of any size or histology; “low-risk” cases were defined as solitary adenoma <1 cm in greatest diameter with tubular or unspecified histology. Multiple adenoma was defined as ≥2 adenomas, consistent with previous studies (12, 13).
According to the distribution of the PGE-M in controls, we categorized PGE-M levels into quartiles and estimated the ORs and the 95% confidence intervals (CI) for colorectal adenoma using conditional logistic regression. Tests for trend were conducted using midpoint of each quartile as a continuous term in the regression models. We obtained similar results using conditional logistic regression models or unconditional logistic regression models with adjustment for matching factors; thus, we present the results from unconditional logistic regression. To evaluate the possible nonlinear shape of the association between PGE-M levels and high-risk adenoma, we fit a restricted cubic spline function with four knots (17). Tests for nonlinearity used the likelihood ratio test, comparing the model with only the linear term with the model with the linear and cubic spline terms.
In multivariate analysis, we adjusted for potential confounders, including regular aspirin/NSAID use [defined as ≥2 standard (325 mg) tablets of aspirin or ≥2 tablets of NSAIDs per week], regular use of multivitamins, red meat intake, pack-years of smoking, body mass index (BMI), physical activity in metabolic equivalent task score hours (MET) per week, menopausal hormone use, alcohol consumption, and energy-adjusted intake of calcium and folate. Missing information was carried forward from available information on prior questionnaires.
We first examined the associations between urinary PGE-M and high-risk and low-risk adenoma groups as defined above. We also compared the association between PGE-M and subtypes of adenoma using a polytomous logistic regression model in which effect estimates were allowed to vary between the subtypes but all covariates were held constant. To calculate a P value for heterogeneity in the relationship between PGE-M and each subtype, we used a likelihood ratio test comparing the polytomous model with a model in which the associations were held constant between case groups.
We also conducted stratified analyses according to subgroups defined by BMI, smoking, use of aspirin or NSAIDs, and calcium levels. To test for multiplicative interaction between stratification factors and PGE-M, we included cross-product terms for stratification factors and PGE-M in our models. We also stratified the cohort according to baseline PGE-M level and examined the association of aspirin and NSAID use at the time of urine collection with risk of adenoma. In addition, we examined the association of acetaminophen use at the time of urine collection with risk of adenoma.
We used SAS version 9.3 (SAS Institute, Inc.) for all analyses with the exception of the polytomous logistic regression model, for which we used Stata version 11.0 (StataCorp). All statistical tests were two-sided and P < 0.05 was considered statistically significant.
Table 1 shows the baseline characteristics of the 420 case participants with either low-risk adenoma (solitary adenoma < 1 cm in greatest diameter with tubular or unspecified histology) or high-risk adenoma [adenoma ≥1 cm in greatest diameter and/or tubulovillous, villous or high-grade dysplasia histology or multiple (≥2) adenomas of any size or histology] and the 420 matched endoscopy-negative control participants at the time of urine collection. The mean age of the study cohort was 67 years. Compared with endoscopy-negative controls, case women in the high-risk adenoma group were more likely to have a higher BMI and less likely to be regularly using aspirin or NSAIDs at the time of urine collection. The median levels of baseline urinary PGE-M in high-risk cases were 6.26 ng/mg Cr. compared with 5.57 ng/mg Cr. in controls (P = <0.001). Among controls, there was no significant difference in levels of urinary PGE-M according to use of aspirin/NSAIDs at the time of urine collection (P = 0.43).
The Spearman correlation coefficients between urinary PGE-M and several lifestyle factors are presented in Table 2. PGE-M directly correlated with age, BMI, and smoking. Table 3 shows the association between urinary PGE-M and the risk of overall colorectal adenoma according to quartile categories determined by the distribution of PGE-M among controls. Compared with the women in the lowest quartile of urinary PGE-M, the multivariate OR for any adenoma was 1.40 (95% CI, 0.92–2.14) for women in the highest quartile (P for linear trend = 0.26). Compared with the women in the lowest quartile of urinary PGE-M, the multivariate OR for high-risk adenoma was 1.66 (95% CI, 1.04–2.67) for women in the highest quartile (P for linear trend = 0.04). However, PGE-M level was not significantly associated with the low-risk adenoma (P for linear trend = 0.30; Table 3). We also used restricted cubic splines to examine the shape of the relationship between PGE-M levels and the risk of high-risk adenoma (Fig. 1). A test for overall significance of the curve was P = 0.04 and the test for a nonlinear relation was P = 0.09.
We also conducted analyses according to subtypes of adenoma defined by histology, size, or number (Supplementary Table S1). Compared with women in the lowest quartile of urinary PGE-M, women in the highest quartile had a multivariate ORs of 1.65 (95% CI, 1.02–2.69) for adenoma with advanced histology (tubulovillous, villous, or high-grade dysplasia); 1.69 (95% CI, 1.00–2.83) for large adenoma (≥1 cm in greatest diameter); and 2.26 (95% CI, 1.25–4.06) for multiple (≥2) adenoma. In contrast, PGE-M level was not significantly associated with early histology (tubular/unspecified), small (<1cm in greatest diameter), or solitary adenoma.
We also examined the association of urinary PGE-M and risk of high-risk adenoma according to strata of selected lifestyle factors associated with adenoma risk (Table 4). The association between PGE-M and high-risk adenoma did not seem to vary according to subgroups defined by regular use of aspirin or NSAID, calcium intake, smoking status, or BMI (P for interaction >0.05 for all factors).
On the basis of the shape of the relationship between PGE-M level and risk of adenoma, we conducted an exploratory analysis in which we stratified the cohort according to high baseline PGE-M level, defined as quartiles 2–4, versus low baseline PGE-M, defined as quartile 1. The association of regular use of anti-inflammatory drugs (≥2 tablets of aspirin or NSAIDs per week) and risk of adenoma seemed to differ according to baseline PGE-M levels (Table 5). Among women with high baseline PGE-M, regular use of anti-inflammatory drugs was associated with a significant reduction in adenoma risk (multivariate OR, 0.61; 95% CI, 0.43–0.87). In contrast, among women with low baseline PGE-M, anti-inflammatory drugs were not associated with a lower risk of adenoma (multivariate OR, 1.05; 95% CI, 0.50–2.19). Regular use of acetaminophen did not affect adenoma risk.
In this large prospective nested case–control study of women, we observed an association between prediagnostic urinary PGE-M and risk of high-risk colorectal adenoma. In contrast, PGE-M was not associated with low-risk adenoma. Moreover, regular use of aspirin or NSAIDs seemed to reduce risk of adenoma risk among women with high, but not low, PGE-M levels. Our results support the potential for urinary PGE-M to serve as a biomarker to define subsets of the population who may obtain differential chemopreventive benefit from anti-inflammatory drugs.
The proinflammatory enzyme PTGS2, also known as COX-2, has been established to play a key role in colorectal carcinogenesis (6). The generation of PGE2 is thought to be the primary mechanism by which PTGS2 promotes the development and progression of neoplasia (8). Animal studies have previously shown that PGE2 administration reverses NSAID-induced adenoma regression and enhances carcinogen-induced tumor incidence (18, 19). In addition, deletion of the PGE2 receptors has shown to result in resistance to the formation of crypt foci, polyps, and cancers (20, 21). A loss of expression of 15-PGDH, an enzyme that degrades PGE2, has also been noted in colorectal neoplasia (8). Furthermore, genetic deletion of microsomal PGE2 synthase-1, an enzyme responsible for PGE2 production, suppresses intestinal carcinogenesis in APC-mutant mouse models (22).
Our results agree with prior studies that have shown that PGE-M, the major urinary metabolite of PGE2, is associated, in a cross-sectional study, with advanced and multiple adenoma as well as colorectal cancer (11–13). PGE-M may also be associated with risk and progression of malignancies other than colon cancer (23). Findings from the Shanghai Women's Health Study cohort suggest that urinary PGE-M may be associated with gastric cancer risk (24). Morris and colleagues found, in a cross-sectional study, that urinary PGE-M is positively associated with lung metastases in patients with breast cancer (25) A prospective study also showed that PGE-M is associated with breast cancer risk in postmenopausal women, although this association was limited to those who did not use NSAIDs (26).
Our study extends these prior results by demonstrating that the association of aspirin and NSAIDs and risk of adenoma seems to differ according to baseline PGE-M levels. These findings support the hypothesis that aspirin and NSAIDs reduce risk of colorectal neoplasia through anti-inflammatory pathways and suggest the possibility that PGE-M may be a noninvasive biomarker by which individuals who may obtain greater chemopreventative benefit from the use of anti-inflammatory drugs can be identified.
Our study has several strengths, including prospective collection of urine before endoscopy and adenoma diagnosis and detailed information on covariates that allowed us to adjust for potential confounding. Moreover, the availability of detailed aspirin and NSAID data allowed the evaluation of whether PGE-M levels significantly modified the effect of these established chemopreventive agents on adenoma risk. Limitations of our study include the inclusion of only female health professionals and the use of a single measurement of PGE-M to reflect overall prostaglandin status. However, measurements of urinary PGE-M taken 1 year apart have been shown to have high intraindividual stability (27). Second, although we show that the lower risk of adenoma associated with aspirin or NSAIDs is most pronounced among individuals with high baseline PGE-M levels, we cannot determine if aspirin or NSAIDs directly reduce PGE-M because we only had a single baseline assessment of PGE-M. Future studies are needed in which changes in PGE-M are assessed based on urine collected before and after aspirin treatment. Third, we had a limited sample size within which to examine the association of PGE-M with adenoma according to subgroups. However, we did not observe statistically significant heterogeneity according to subgroups and generally observed similar associations between PGE-M and high-risk adenoma within each subgroup. Last, our participants did not all undergo a clearing endoscopy before urine collection. Thus, PGE-M levels may be elevated before the initiation of an adenoma or reflect progression of preexisting lesions.
In summary, we have shown that urinary PGE-M is associated with an increased risk for advanced, large, and multiple adenoma. In addition, aspirin or NSAID use seemed to be more strongly associated with lower risk of adenoma among individuals with high PGE-M levels compared with those with low PGE-M levels. These results provide proof-of-principle of the mechanistic importance of inflammation, PTGS-2, and prostaglandin pathways in colorectal carcinogenesis. In addition, although aspirin and NSAIDs have been shown to be effective chemopreventives in colorectal neoplasia, widespread use of these agents has been hindered due to concerns about its safety profile. Thus, further investigation of urinary PGE-M as a biomarker to risk-stratify patients for chemopreventive response to anti-inflammatory drugs is a high priority.
Disclosure of Potential Conflicts of Interest
A.T. Chan is a consultant/advisory board member of Bayer Healthcare and Pozen, Inc. No potential conflicts of interest were disclosed by the other authors.
Certain data used in this publication were obtained from the DPH. The authors assume full responsibility for analyses and interpretation of these data.
Conception and design: C.S. Fuchs, A.T. Chan
Development of methodology: C.S. Fuchs, E.L. Giovannucci, A.T. Chan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G.L. Milne, C.S. Fuchs, A.T. Chan
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N. Bezawada, M. Song, K. Wu, R S. Mehta, S. Ogino, C.S. Fuchs, E.L. Giovannucci, A.T. Chan
Writing, review, and/or revision of the manuscript: N. Bezawada, M. Song, K. Wu, S. Ogino, C.S. Fuchs, E.L. Giovannucci, A.T. Chan
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Song, C.S. Fuchs, A.T. Chan
Study supervision: A.T. Chan
This work was supported by U.S. NIH grants P01 CA55075 (to S. Hankinson), UM1CA167552 (W. Willett), P50CA127003 (to C.S. Fuchs), P01CA087969 (to E.L. Giovannucci), R01CA151993 (to S. Ogino), R01 CA137178 (to A.T. Chan), and K24 DK098311 (to A.T. Chan). A. Chan is a Damon Runyon Cancer Research Foundation Clinical Investigator.
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.
The authors thank the participants and staff of the NHS for their valuable contributions as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, and WY. In addition, this study was approved by the Connecticut Department of Public Health (DPH) Human Investigations Committee.
Note: Supplementary data for this article are available at Cancer Prevention Research Online (http://cancerprevres.aacrjournals.org/).
- Received April 11, 2014.
- Revision received May 6, 2014.
- Accepted May 6, 2014.
- ©2014 American Association for Cancer Research.