Prediagnostic Serum Levels of Fatty Acid Metabolites and Risk of Ovarian Cancer in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial

Published OnlineFirst September 27, 2018; DOI: 10.1158/1055-9965.EPI-18-0392

Research Article Cancer Epidemiology, Biomarkers Prediagnostic Serum Levels of Fatty Acid & Prevention Metabolites and Risk of Ovarian Cancer in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial Manila Hada1, Matthew L. Edin2, Patricia Hartge1, Fred B. Lih2, Nicolas Wentzensen1, Darryl C. Zeldin2, and Britton Trabert1

Abstract

Background: Evidence suggests that inflammation [2.47 (1.32–4.60), 0.005], 9-HODE [1.97 (1.06–3.68), increases risk for ovarian cancer. Aspirin has been shown to 0.03], 9,12,13-THOME [2.25 (1.20–4.21), 0.01]. In analyses decrease ovarian cancer risk, though the mechanism is by subtype, heterogeneity was suggested for 8-HETE [serous unknown. Studies of inflammatory markers, lipid molecules OR (95% CI): 2.53 (1.18–5.39) vs. nonserous OR (95% CI): such as arachidonic acid, linoleic acid, and alpha-linoleic acid 1.15 (0.56–2.36), Phet 0.1] and 12,13-EpOME [1.95 (0.90– metabolites, and development of ovarian cancer are essential 4.22) vs. 0.82 (0.39–1.73), 0.05]. to understand the potential mechanisms. Conclusions: Women with increased levels of five fatty Methods: We conducted a nested case–control study (157 acid metabolites (8-HETE, 12,13-DHOME, 13-HODE, cases/156 matched controls) within the Prostate, Lung, Colorec- 9-HODE, and 9,12,13-THOME) were at increased risk of tal, and Ovarian (PLCO) Cancer Screening Trial. Unconditional developing ovarian cancer in the ensuing decade. All five logistic regression was used to estimate the association between metabolites are derived from either arachidonic acid prediagnostic serum levels of 31 arachidonic acid/linoleic acid/ (8-HETE) or linoleic acid (12,13-DHOME, 13-HODE, alpha-linoleic acid metabolites and risk of ovarian cancer. 9-HODE, 9,12,13-THOME) via metabolism through the Results: Five of the 31 arachidonic acid/linoleic acid/alpha- LOX/cytochrome P450 pathway. linoleic acid (free fatty acids) metabolites were positively Impact: The identification of these risk-related fatty acid associated with ovarian cancer risk: 8-HETE [tertile 3 vs. 1: metabolites provides mechanistic insights into the etiology OR 2.53 (95% confidence interval [CI] 1.18–5.39), Ptrend of ovarian cancer and indicates the direction for future 0.02], 12,13-DHOME [2.49 (1.29–4.81), 0.01], 13-HODE research.

Introduction risk (6). Although a specific mechanism is yet to be determined, reduced risk of ovarian cancer due to aspirin might be due to the Inflammation is thought to play a role in the development and medication's anti-inflammatory effect. Aspirin and other NSAIDs progression of ovarian cancer. Usually inflammation is beneficial block the synthesis of proinflammatory prostaglandin synthesis by such that it activates the immune process and protects the body inhibiting the COX enzyme (7). from infections and diseases, but chronic inflammation induces In addition to prostaglandin synthesis, the inflammatory prolonged exposure of the cells to the mediators of inflammation response involves biologically active lipid molecules, arachidonic (1, 2). These mediators can then promote tumorigenesis. Inflam- acid, and an essential fatty acid linoleic acid and its metabolites mation during ovulation or inflammatory chronic disease that (Fig. 1; refs. 8–10). Essential fatty acids mainly comprise two affects the fallopian tubes and/or ovary (e.g., endometriosis and groups: omega-6 fatty acids (linoleic acid) and omega-3 fatty acids pelvic inflammatory disease) is associated with increased risk of (alpha-linoleic acid). Omega-6 fatty acids are converted to ara- ovarian cancer (3–5). The use of nonsteroidal anti-inflammatory chidonic acid, and omega-3 fatty acids are converted to docosa- drugs (NSAIDs), such as aspirin, is associated with reductions in hexaenoic acid and eventually to arachidonic acid (11, 12). In addition to essential fatty acid metabolism to arachidonic acid, 1Division of Cancer Epidemiology and Genetics, National Cancer Institute, arachidonic acid is also released from the phospholipids of the National Institutes of Health, Bethesda, Maryland. 2National Institute of Envi- cell membrane during inflammation in response to various ronmental Health Sciences, National Institutes of Health, Research Triangle Park, stimuli like cytokines, hormones, and stress (7). Arachidonic acid North Carolina. is acted upon by three enzymes: COX, lipoxygenase (LOX), and Note: Supplementary data for this article are available at Cancer Epidemiology, cytochrome P450. The COX activity on arachidonic acid produces Biomarkers & Prevention Online (http://cebp.aacrjournals.org/). PGG2 and PGH2, which is later converted to prostanoids includ- Corresponding Author: Manila Hada, National Cancer Institute, 9609 Medical ing prostaglandins (PGD2, PGE2, and PGF2a), prostacyclins Center, Rockville, MD 20850; Phone: 240-276-7603; E-mail: (PG12), and thromboxanes (TXA2 and TXB2). COX exists in [email protected] two isoforms, COX-1 and COX-2. Three forms of LOXs (5-LOX, doi: 10.1158/1055-9965.EPI-18-0392 12-LOX, and 15-LOX) acts on arachidonic acid to produce 2018 American Association for Cancer Research. hydroxyperoxyeicosatetraenoic acids (HPETE), HETEs, and

www.aacrjournals.org 189

Hada et al.

Dietary Source Cell Membrane

8-iso-PGF2α Non-enzymatic Alpha-Linoleic Acid Linoleic Acid (LA) Arachidonic Acid (AA) (ALA) 11-HETE Cox/ 15LOX CYP450 Nonenzymatic Eicosapentaenoic COX LOX CYP450 Acid 19,20-EpDPE 19,20-EpDPE

PGG 15LOX 12LOX 8LOX 5LOX 17,18-EpETE 9,10-EpOME 12,13-EpOME 2

Docosahexaenoic 9-HpODE 13-HODE 15-HETE 12-HETE 8-HETE 5-HpETE 20-HETE* 5,6-EET** 8,9-EET 11,12-EET 14,15-EET Acid 17,18-DiHETE 9,10-DHOME 12,13-DHOME PGH CytP450 2 9-HODE 9,12,13-THOME 19-HETE* 5,6-DHET 8,9-DHET 11,12-DHET 14,15-DHET 19,20-EpDPE 5-HETE LTA

9,10,13-THOME PGI PGF PGE PGD TXA 19,20-DiHDPA 2 2α 2 2 2 LTB4

6-keto-PGF1α TXB2

Figure 1. Essential fatty acid metabolism pathway. The current assay measured the 31 arachidonic acid/linoleic acid/alpha-linoleic acid metabolites in the shaded boxes. The metabolites in the outlined boxes were not measured using the current assay. , 20-HETE and 19-HETE metabolites undetectable in nearly all samples. , 5,6-EET was excluded from analyses given its known limitations in its detection via LC/MS-MS.

leukotrienes (LT). The enzyme cytochrome P450 produces HETEs collection with follow-up through December 2010. To ensure a and epoxides. Free fatty acid (arachidonic acid/linoleic acid/ relatively equal distribution of specimens between 2 and 14 years alpha-linoleic acid) metabolites are implicated in various signal- prior to diagnosis, 10.9% of samples selected were measured at ing pathways involved in physiologic processes such as cell baseline and the remaining at follow-up visits (16.3% year 1, differentiation, cell migration, platelet aggregation, angiogenesis, 25.9% year 2, 13.7% year 3, and 33.2% year 4). For inclusion in and regulation of immune functions (8, 13), and thus may play a our study, serum samples were selected from a prediagnostic blood role in cancer initiation and/or promotion. draw given the availability of unthawed serum, consent to bio- Because of a potentially tumorigenic role of arachidonic acid, chemical studies, completion of the baseline questionnaire, and no the inhibitors of arachidonic acid metabolites have been widely history of cancer (except nonmelanoma skin cancer). There were studied for their potential role in cancer treatment but the precise 160 cases identified out of which three were excluded from the molecular mechanism by which these metabolites drive tumor- study because the sample volume was not sufficient. Controls were igenesis is unknown (14). Epidemiologic studies investigating the women with no history of oophorectomy at the time of diagnosis role of free fatty acid metabolites in ovarian cancer are limited. To of their matched case and were matched on age at blood collection understand the role of arachidonic acid and linoleic acid pathway- (55–59, 60–64, 65–69, 70þ years), race (white, black, other), study induced inflammation in the etiology of ovarian cancer, we center, and time (a.m. and p.m.) and date (3-month categories) of assessed several free fatty acid metabolites in a nested case–control blood collection. Serum specimens from a single visit were mea- study within the Prostate, Lung, Colorectal, and Ovarian (PLCO) sured for each study subject. All the study participants provided Cancer Screening Trial. As ovarian cancer subtypes are known to written informed consent and the study received approval from the have a heterogeneous etiology, we further evaluated associations institutional review board of the involved study centers and from by serous/nonserous subtype and time between blood draw and the National Cancer Institute. cancer diagnosis. Laboratory analysis Levels of 34 free fatty acid metabolites generated through three Materials and Methods different pathways—COX pathway, cytochrome P450 pathway, Study design and LOX pathway—were measured in serum. All free fatty acid We conducted a nested case–control study within the screening metabolites were separated by electrospray ionization LC (Agilent arm of the PLCO Cancer Screening Trial, a randomized two-arm 1200 series capillary HPLC) and quantified using tandem mass screening trial of men and women, ranging in age 55 to 74 years spectrometry (MS/MS, MDS Sciex API 3000, negative ion mode) was conducted between 1993 and 2001 (15). Briefly, 78,216 with a Turbolon Spray and scheduled multiple reaction. The women recruited from 10 centers across the United States details of the assay have been published previously (17–19). (Birmingham, AL; Denver, CO; Detroit, MI; Honolulu, HI; Concentrations (pg/mL) were quantified with Analyst software Marshfield, WI; Minneapolis, MN; Pittsburgh, PA; Salt Lake City, (v 1.5; Applied Biosystem) using metabolite and internal standard UT; St. Louis, MO; and Washington, DC). Participants in the peaks for each sample. screening arm provided a blood sample at baseline and during five To evaluate assay performance, we included duplicate quality follow-up medical examinations and were stored at 70 C (16). control (QC) samples in each batch and calculated the within- All participants completed a self-administered baseline question- batch and between-batch coefficients of variation (CV). We also naire. Cancer cases were identified by annual mailed question- calculated the intraclass correlation coefficient (ICC) for each naires which were then confirmed by linking to population-based analyte. 51.6% of CVs were <10%, 25.8% were 10% to 15%, and registries and the National Death Index. Medical and pathologic 22.5% were 15% to 20% (Supplementary Table S1). We excluded records were obtained when possible. from evaluation metabolite 5,6-EET given known limitations in To avoid the possibility of reverse causality, cases for our study its detection via LC/MS-MS, we further excluded two metabolites were 157 individuals diagnosed between 2 and 14 years after blood (19-HETE and 20-HETE) that were undetectable in nearly all

190 Cancer Epidemiol Biomarkers Prev; 28(1) January 2019 Cancer Epidemiology, Biomarkers & Prevention

Serum Fatty Acid Metabolites and Ovarian Cancer Risk

samples. Thus, after exclusions, 31 metabolites were evaluated Table 1. Characteristics of ovarian cancer cases and controls in a nested case– (17, 20, 21). control study in the PLCO Cancer Screening Trial Control Ovarian cancer Statistical analysis (N ¼ 156) cases (N ¼ 157) a a We used generalized linear models adjusted for age to calculate N (%) N (%) Age at randomization geometric mean concentrations of free fatty acid metabolites by – 59 years 46 (29.5) 46 (29.3) case control status and by other factors (e.g., smokers vs. non- 60–64 years 46 (29.5) 46 (29.3) smokers, normal BMI versus overweight/obese BMI, and aspirin/ 65–69 years 40 (25.6) 41 (26.1) Ibuprofen users vs. nonusers). P-values were calculated using a 70 years 24 (15.4) 24 (15.3) Wald test. Race We used unconditional logistic regression models to calculate Non-Hispanic white 144 (92.3) 144 (91.7) odds ratios and 95% confidence intervals (CI) for the association Non-Hispanic black, Hispanic, Asian 12 (7.7) 13 (8.3) Highest education level attained between prediagnostic serum levels of free fatty acid metabolites High school or less 49 (31.4) 45 (29.0) and risk of ovarian cancer. All models were adjusted for matching Some post-high school training 52 (33.3) 55 (35.0) factors and potential confounders: parity (nulliparous/parous), College graduate 55 (35.1) 57 (36.3) duration of oral contraceptive use (never, 1–5 years, 6þ years), BMI 2 duration of menopausal hormone therapy use (never, 1–5 years, <25 kg/m 66 (42.3) 66 (42.6) – 2 6þ years), cigarette smoking status (never, former, current), and 25 29.9 kg/m 54 (35.0) 56 (36.1) 2 – þ 2 30 kg/m 36 (23.1) 33 (21.3) body mass index (BMI; <25, 25 29.9, 30 kg/m ). The risk of Cigarette smoking status ovarian cancer was examined across tertile categories of the Never 101 (64.7) 84 (53.5) metabolites based on the distribution among controls. Metabo- Former 17 (10.9) 12 (7.6) lites with 90% or more individuals with nondetectable levels were Current 38 (24.3) 61 (38.8) categorized into two groups (detectable vs. undetectable). A Wald Age at menarche test using tertile categories as a continuous variable was used to <11 years 27 (17.1) 34 (21.7) 12–13 years 80 (51.3) 89 (56.7) evaluate trend. For markers associated with ovarian cancer risk, we 13þ years 49 (31.2) 34 (21.7) fl further adjusted for the in ammatory markers (CRP, TNFa, and Parity IL8) that were positively associated with ovarian cancer risk in a Nulliparous 149 (95.0) 146 (93.0) previous analysis (22). Parous 8 (5.1) 11 (7.0) We also explored the association between free fatty acid meta- Duration of oral contraceptive use bolites and risk of ovarian cancer by histologic subtypes (serous Never or <1 year 77 (49.4) 86 (54.8) 1–5 years 50 (32.0) 48 (30.6) vs. nonserous) and number of years from blood collection to 6þ years 29 (18.6) 23 (14.6) – cancer diagnosis (cancer diagnosis 2 5 years vs. >5 years from Regularly uses aspirin or ibuprofen blood draw). P-values for heterogeneity were based on the Wald Does not use aspirin or ibuprofen 66 (43.0) 55 (35.0) test for the ordinal metabolite variable from a case-only model Does not use aspirin but uses ibuprofen 21 (13.5) 25 (16.0) with serous histology or the shortest time (2–5 years) between Uses aspirin but not ibuprofen 44 (28.4) 53 (34.0) blood draw and diagnosis as the reference group. Uses both aspirin and ibuprofen 24 (15.5) 24 (15.4) Duration of menopausal hormone therapy use We used fixed effect meta-analysis to calculate summary objec- Never or <1 year 63 (41.0) 45 (28.7) tive responses (OR) of the association of fatty acid metabolite 1–5 years 48 (30.7) 50 (32.0) pathways (Fig. 1) with the risk of ovarian cancer overall and by 6þ years 45 (28.8) 62 (39.5) histologic subtypes. aValues may not sum to total because of missing data. We evaluated effect modification by BMI (normal vs. overweight/obese) and use of medication (aspirin/ibuprofen users vs. fi aspirin/ibuprofen nonusers) using strati ed models. Likelihood longer durations (6þ years) than controls (39.5% in cases vs. ratio tests were calculated to test for interaction using the cross- 28.8% in controls). The median number of years from blood draw product term. Because of a limited number of current smokers, we to the year of cancer diagnosis was 8.2 years [interquartile range fi could not evaluate effect modi cation by smoking status, but we (IQR): 5.2–10.8]. conducted a sensitivity analysis restricted to never/former smo- The age adjusted geometric means of the free fatty acid meta- kers. As a secondary analysis, we evaluated the association bolites were similar between cases and control (Table 2). Out of between free fatty acid metabolites and the risk of ovarian cancer 31 metabolites analyzed for an association with ovarian cancer, adjusting for aspirin use. five (8-HETE, 12,13-DHOME, 13-HODE, 9-HODE, and 9,12,13- To account for multiple comparisons, we applied the false THOME) were positively associated with ovarian cancer risk discovery rate (FDR) for the primary free fatty acid metabolites (Table 3): 8-HETE [tertile 3 vs. 1: OR 2.53 (95% CI, 1.18– associations with the ovarian cancer. All other analyses were 5.39), Ptrend 0.02], 12,13-DHOME [2.49 (95% CI, 1.29–4.81), considered exploratory and not corrected for multiple compar- 0.01], 13-HODE [2.47 (1.32–4.60), 0.005], 9-HODE [1.97 fi isons. All tests were two-sided and statistical signi cance was (1.06–3.68), 0.03], 9,12,13-THOME [2.25 (1.20–4.21), 0.01]. fi P de ned using a -value <0.05. Analyses were conducted in SAS After accounting for multiple comparison, three metabolites version 9.4 (SAS Inc.). (12,13-DHOME, 13-HODE, and 9,12,13-THOME) remained associated with ovarian cancer risk at FDR < 0.10. Increased Results ovarian cancer risk with the five metabolites remained after The participants for the study were predominantly white adjusting for the potential mediating effects of other inflamma- (92.3%; Table 1). Cases used menopausal hormonal therapy for tory markers (e.g., CRP, TNFa, IL8; Supplementary Table S2).

www.aacrjournals.org Cancer Epidemiol Biomarkers Prev; 28(1) January 2019 191

Hada et al.

Table 2. Age-adjusted GM of arachidonic acid and linoleic acid–derived metabolites for ovarian cancer cases and controls in a nested case–control study in the PLCO Cancer Screening Trial Control Ovarian cancer (N ¼ 156) cases (N ¼ 157) Metabolites (pg/ml) GM (95% CI) GM (95% CI) P-value Alpha-linoleic acid derived 17-18-Epoxy-eicosatrienoic acids (17,18-EpETE)a 827 (211–3,235) 349 (65–1,862) 0.40 17-18-Dihydroxyeicosatetraenoic acids (17,18-DiHETE) 1,759 (1,621–1,908) 1,850 (1,705–2,006) 0.39 19-20-Epoxy-docosapentaenoic acid (19,20-EpDPE) 3,044 (2,784–3,327) 3,193 (2,927–3,482) 0.45 19-20-Dihydroxy-docosapentaenoic acid (19,20-DiHDPA) 1,140 (1,055–1,232) 1,190 (1,102–1,286) 0.44 Linoleic acid/COX derived 9-Hydroxyoctadecadienoic acid (9-HODE) 5,988 (5,423–6,612) 6,468 (5,859–7,140) 0.28 Trihydroxyoctadecenoic acids (9,10,13-THOME) 2,685 (2,470–2,918) 2,929 (2,696–3,183) 0.15 Linoleic acid/LOX derived 13-Hydroxyoctadecadienoic acid (13-HODE) 6,018 (5,479–6,610) 6,460 (5,883–7,093) 0.29 Trihydroxyoctadecenoic acids (9,12,13-THOME) 1,343 (1,204–1,499) 1,442 (1,293–1,608) 0.37 Linoleic acid/P450 derived 9-10-Epoxy-octadecenoic acid (9,10-EpOME) 2,016 (1,092–3,723) 1,426 (774–2,629) 0.43 9-10-Dihydroxyoctadecenoic acid (9,10-DHOME) 4,593 (4,158–5,075) 4,901 (4,437–5,412) 0.37 12–13-Epoxy-octadecenoic acid (12,13-EpOME)) 32,993 (29,346–37,092) 35,060 (31,185–39,416) 0.47 12-13-Dihydroxyoctadecenoic acid (12,13-DHOME) 5,041 (4,529–5,612) 5,614 (5,045–6,247) 0.16 Arachidonic acid/COX derived

6-keto-Prostaglandin F1a (6-keto-PGF1a) 1,453 (1,300–1,625) 1,535 (1,373–1,716) 0.50 Prostaglandin F2a (PGF2a) 1,367 (1,174–1,591) 1,492 (1,284–1,735) 0.42 Prostaglandin E2 (PGE2) 72.5 (59.1–89.0) 72.7 (59.2–89.2) 0.99 a Prostaglandin D2 (PGD2) 249 (180–343) 211 (153–291) 0.48 Thromboxane B2 (TXB2) 278 (191–404) 275 (189–398) 0.96 Arachidonic acid/LOX derived 15-Hydroxyeicosatetraenoic acids (15-HETE) 384 (347–425) 390 (352–431) 0.83 12-Hydroxyeicosatetraenoic acids (12-HETE) 4,844 (3,932–5,966) 4,941 (4,014–6,082) 0.89 8-Hydroxyeicosatetraenoic acids (8-HETE) 5,022 (4,551–5,541) 5,325 (4,828–5,874) 0.41 5-Hydroxyeicosatetraenoic acids (5-HETE) 661 (615–710) 682 (635–733) 0.54 Leukotriene B4 (LTB-4) 200 (169–236) 255 (216–300) 0.04 Arachidonic acid/P450 derived 5–6-Dihydroxyeicosatrienoic acids (5,6-DHET) 348 (322–376) 352 (326–381) 0.84 8-9-Epoxyeicosatrienoic acids (8,9-EET) 668 (612–729) 670 (614–731) 0.96 8-9-Dihydroxyeicosatrienoic acids (8,9-DHET) 226 (214–240) 230 (218–244) 0.66 11-12-Epoxyeicosatrienoic acids (11,12-EET) 709 (638–788) 685 (617–761) 0.65 11-12-Dihydroxyeicosatrienoic acids (11,12-DHET) 392 (370–414) 403 (381–426) 0.48 14-15-Epoxyeicosatrienoic acids (14,15-EET) 1,696 (1,511–1,905) 1,805 (1,605–2,029) 0.46 14-15-Dihydroxyeicosatrienoic acids (14,15-DHET) 523 (496–551) 523 (497–551) 0.98 Arachidonic acid/nonenzymatic derived 11-Hydroxyeicosatetraenoic acids (11-HETE) 190 (167–215) 191 (168–216) 0.94 a 8-Iso-Prostaglandin F2a (8-iso-PGF2a) 718 (392–1,317) 571 (302–1,080) 0.59 GM, geometric mean. a PGD2, 8-iso-PGF2a, and 17,18-EpETE had 90% of the samples below the detection limit. The GM and CI presented are for detectable samples only.

In analyses by subtype (Table 3), heterogeneity was suggested nonserous ovarian cancers. The alpha-linoleic acid–derived [0.95 for 8-HETE [serous OR (95% CI), 2.53 (1.18–5.39) vs. nonserous (0.76–1.19), 0.66 vs. 1.27 (1.01–1.59), 0.04)] pathway was OR (95% CI): 1.15 (0.56–2.36), Phet 0.10] and 12,13-EpOME elevated for nonserous ovarian cancers but not for serous ovarian [1.95 (0.90–4.22) vs. 0.82 (0.39–1.73), Phet 0.05]. cancers. In pathway analyses, the linoleic acid/COX–derived [summary In sensitivity analyses, free fatty acid metabolites associations OR (95% CI), P-value: 0.95 (0.76,1.19), 0.004], linoleic acid/ did not vary by time between blood draw and diagnosis (Sup- LOX–derived [1.52 (1.22–1.90), <0.001], and linoleic acid/cyto- plementary Table S3). The results were not significantly modified chrome P450–derived [(1.21 (1.04,1.41), 0.02] pathways were by BMI (normal vs. overweight/obese; Supplementary Table S4). associated with increased risk of ovarian cancer (Fig. 2). In analyses by subtype (Supplementary Fig. S1), linoleic acid/ COX–derived [serous OR (95% CI), P-value: 1.38 (1.04–1.84), Discussion 0.03 vs. nonserous OR (95% CI), P-value: 1.31(0.99–1.73), 0.06] In this prospective study exploring the association between free and linoleic acid/LOX–derived [1.49 (1.13–1.97), 0.005 vs. 1.54 fatty acids and ovarian cancer, we report that five out of 31 (1.16–2.04), 0.003)]–pathways were associated with increased metabolites (8-HETE, 12,13-DHOME, 13-HODE, 9-HODE, and risk of both serous and nonserous ovarian cancers. Arachidonic 9,12,13-THOME) were associated with increased ovarian cancer acid/LOX–derived [1.31 (1.11–1.54), 0.001 vs. 1.11 (0.94–1.32), risk. 8-HETE is produced by the metabolism of arachidonic acid 0.22)] and arachidonic acid/cytochrome P450–derived [1.18 and four metabolites: 12,13-DHOME, 13-HODE, 9-HODE, and (1.02–1.36), 0.03 vs. 1.00 (0.86–1.15), 0.95)] pathways were 9,12,13-THOME, are derived from linoleic acid oxidation. Further associated with increased risk of serous ovarian cancers but not supporting the individual metabolite associations, we observed

192 Cancer Epidemiol Biomarkers Prev; 28(1) January 2019 Cancer Epidemiology, Biomarkers & Prevention

Serum Fatty Acid Metabolites and Ovarian Cancer Risk

Table 3. Association of arachidonic acid and linoleic acid–derived metabolites and risk of overall ovarian cancer and histologic subtypes (serous and nonserous) of ovarian cancer in a nested case–control study in the PLCO Cancer Screening Trial Control Cases Overall cases Serous cases Nonserous cases Metabolites (pg/ml) N (%) N (%) ORa (95% CI)a N (%) OR (95% CI) N (%) OR (95% CI) P-Hetb Alpha-linoleic acid derived 17,18-EpETEc 0 152 (97.4) 154 (98.1) 1.00 (Reference) 83 (100) 1.00 (Reference) 63 (95.5) 1.00 (Reference) >0 (detectable) 4 (2.6) 3 (1.9) 0.91 (0.19–4.42) —— 3 (4.5) 1.96 (0.39–9.92) 17,18-DiHETEd 1,356.67 52 (33.3) 41 (26.1) 1.00 (Reference) 25 (30.1) 1.00 (Reference) 16 (24.2) 1.00 (Reference) >1,356.67–2,070 51 (32.7) 55 (35.0) 1.27 (0.70–2.30) 31 (37.4) 1.04 (0.51–2.14) 22 (33.3) 1.41 (0.64–3.11) >2,070 53 (34.0) 61 (38.9) 1.48 (0.81–2.70) 27 (32.5) 0.96 (0.46–2.03) 28 (42.4) 1.89 (0.86–4.15) 0.55 e Ptrend 0.20 0.63 0.10 19,20-EpDPE 1,900 52 (33.3) 35 (22.3) 1.00 (Reference) 18 (21.7) 1.00 (Reference) 15 (22.7) 1.00 (Reference) >1,900–3,448.33 50 (32.1) 66 (42.0) 1.87 (1.03–3.39) 34 (41.0) 1.67 (0.79–3.52) 31 (47.0) 2.31 (1.07–4.97) >3,448.33 54 (34.6) 56 (35.7) 1.42 (0.76–2.67) 31 (37.4) 1.26 (0.58–2.76) 20 (30.3) 1.43 (0.62–3.31) 0.99 Ptrend 0.32 0.65 0.39 19,20-DiHDPA 883.33 52 (33.3) 44 (28.0) 1.00 (Reference) 27 (32.5) 1.00 (Reference) 17 (25.8) 1.00 (Reference) >883.33–1,390 51 (32.7) 56 (35.7) 1.45 (0.80–2.61) 26 (31.3) 1.04 (0.50–2.17) 27 (40.9) 1.84 (0.86–3.96) >1,390 53 (34.0) 57 (36.3) 1.20 (0.65–2.19) 30 (36.1) 0.86 (0.41–1.81) 22 (33.3) 1.38 (0.62–3.10) 0.77 Ptrend 0.60 0.47 0.42 Linoleic acid/COX derived 9-HODEf 4,383.33 52 (33.3) 41 (26.1) 1.00 (Reference) 18 (21.7) 1.00 (Reference) 22 (33.3) 1.00 (Reference) >4,383.33–7,450 51 (32.7) 48 (30.6) 1.34 (0.72–2.48) 30 (36.1) 1.93 (0.88–4.24) 15 (22.7) 0.81 (0.36–1.83) >7,450 53 (34.0) 68 (43.3) 1.97 (1.06–3.68) 35 (42.2) 2.39 (1.08–5.29) 29 (43.9) 1.42 (0.65–3.08) 0.39 Ptrend 0.03 0.06 0.36 9,10,13-THOME 2,153.33 52 (33.3) 35 (22.3) 1.00 (Reference) 18 (21.7) 1.00 (Reference) 15 (22.7) 1.00 (Reference) >2,153.33–3,070.1 51 (32.7) 62 (39.5) 1.90 (1.02–3.53) 35 (42.2) 1.80 (0.83–3.88) 24 (36.4) 1.95 (0.85–4.45) >3,070.1 53 (34.0) 60 (38.2) 1.83 (1.02–3.66) 30 (36.1) 1.80 (0.81–4.00) 27 (40.9) 2.00 (0.88–4.58) 0.81 Ptrend 0.05 0.22 0.07 Linoleic acid/LOX derived 13-HODEf 4,500 53 (34.0) 36 (22.9) 1.00 (Reference) 17 (20.5) 1.00 (Reference) 17 (25.8) 1.00 (Reference) >4,500–6,800 50 (32.1) 49 (31.2) 1.71 (0.90–3.24) 28 (33.7) 2.11 (0.94–4.73) 21 (31.8) 1.55 (0.68–3.53) >6,800 53 (34.0) 72 (45.9) 2.47 (1.32–4.60) 38 (45.8) 2.96 (1.34–6.56) 28 (42.4) 1.93 (0.86–4.34) 0.51 Ptrend 0.005 0.01 0.12 9,12,13-THOMEf 913.33 52 (33.3) 35 (22.3) 1.00 (Reference) 21 (25.3) 1.00 (Reference) 12 (18.2) 1.00 (Reference) >913.33–1,446.67 51 (32.7) 53 (33.8) 1.61 (0.86–3.02) 29 (34.9) 1.22 (0.57–2.62) 22 (33.3) 2.18 (0.91–5.23) >1,446.67 53 (34.0) 69 (43.9) 2.25 (1.20–4.21) 33 (39.8) 1.68 (0.78–3.61) 32 (48.5) 3.00 (1.28–7.03) 0.29 Ptrend 0.01 0.16 0.01 Linoleic acid/P450 derived 9,10-EpOME 6,950 52 (33.3) 58 (36.9) 1.00 (Reference) 33 (39.8) 1.00 (Reference) 21 (31.8) 1.00 (Reference) >6,950–19,600 51 (32.7) 44 (28.0) 0.77 (0.43–1.40) 15 (18.1) 0.48 (0.22–1.06) 27 (40.9) 1.22 (0.58–2.56) >19,600 53 (34.0) 55 (35.0) 0.94 (0.53–1.65) 35 (42.2) 1.12 (0.57–2.19) 18 (27.3) 0.84 (0.39–1.81) 0.63 Ptrend 0.83 0.70 0.56 9,10-DHOME 2,956.67 52 (33.3) 32 (20.4) 1.00 (Reference) 17 (20.5) 1.00 (Reference) 14 (21.2) 1.00 (Reference) >2,956.67–5,400 51 (32.7) 68 (43.3) 2.15 (1.16–3.99) 42 (50.6) 2.17 (1.02–4.60) 21 (31.8) 1.86 (0.80–4.33) >5,400 53 (34.0) 57 (36.3) 1.90 (0.98–3.67) 24 (28.9) 1.35 (0.58–3.11) 31 (47.0) 2.50 (1.06–5.90) 0.15 Ptrend 0.07 0.60 0.04 12,13-EpOME 23,050 52 (33.3) 46 (29.3) 1.00 (Reference) 17 (20.5) 1.00 (Reference) 27 (40.9) 1.00 (Reference) >23,050–48,000 50 (32.1) 48 (30.6) 1.16 (0.64–2.12) 30 (36.1) 2.23 (1.01–4.89) 16 (24.2) 0.63 (0.29–1.38) >48,000 54 (34.6) 63 (40.1) 1.35 (0.75–2.44) 36 (43.4) 1.95 (0.90–4.22) 23 (34.9) 0.82 (0.39–1.73) 0.05 Ptrend 0.32 0.13 0.67 12,13-DHOMEf 3,226.67 52 (33.3) 30 (19.1) 1.00 (Reference) 16 (19.3) 1.00 (Reference) 13 (19.7) 1.00 (Reference) >3,226.67–6,233.33 51 (32.7) 59 (37.6) 2.08 (1.11–3.88) 34 (41.0) 2.11 (0.97–4.58) 22 (33.3) 1.83 (0.79–4.22) >6,233.33 53 (34.0) 68 (43.3) 2.49 (1.29–4.81) 33 (39.7) 2.15 (0.95–4.87) 31 (47.0) 2.54 (1.07–6.04) 0.52 Ptrend 0.01 0.12 0.04 (Continued on the following page)

www.aacrjournals.org Cancer Epidemiol Biomarkers Prev; 28(1) January 2019 193

Hada et al.

6-keto-PGF1a 1,138.33 52 (33.3) 48 (30.6) 1.00 (Reference) 25 (30.1) 1.00 (Reference) 20 (30.3) 1.00 (Reference) >1,138.33–1,973.33 50 (32.1) 51 (32.5) 1.33 (0.73–2.43) 30 (36.1) 1.58 (0.75–3.31) 19 (28.8) 1.09 (0.49–2.41) >1,973.33 54 (34.6) 58 (36.9) 1.42 (0.77–2.62) 28 (33.7) 1.33 (0.62–2.86) 27 (40.9) 1.47 (0.67–3.24) 0.38 Ptrend 0.26 0.32 0.43

PGF2a 728.33 52 (33.3) 42 (26.8) 1.00 (Reference) 22 (26.5) 1.00 (Reference) 17 (25.8) 1.00 (Reference) >728.33–2,075.00 51 (32.7) 64 (40.8) 1.48 (0.82–2.65) 32 (38.6) 1.43 (0.69–2.95) 28 (42.4) 1.61 (0.75–3.45) >2,075.00 53 (34.0) 51 (32.5) 1.07 (0.58–1.96) 29 (34.9) 1.02 (0.48–2.18) 21 (31.8) 1.24 (0.56–2.75) 0.89 Ptrend 0.86 0.91 0.60 PGE2 43.28 52 (33.3) 56 (35.7) 1.00 (Reference) 31 (37.4) 1.00 (Reference) 20 (30.3) 1.00 (Reference) >43.28–124.67 51 (32.7) 46 (29.3) 0.74 (0.41–1.34) 22 (26.5) 0.56 (0.27–1.19) 22 (33.3) 1.03 (0.47–2.24) >124.67 53 (34.0) 55 (35.0) 0.95 (0.53–1.72) 30 (36.1) 0.86 (0.41–1.81) 24 (36.4) 1.22 (0.56–2.64) 0.52 Ptrend 0.87 0.66 0.63 c PGD2 0 124 (79.5) 125 (79.6) 1.00 (Reference) 68 (81.9) 1.00 (Reference) 51 (77.3) 1.00 (Reference) >0 (detectable) 32 (20.5) 32 (20.4) 1.01 (0.56–1.84) 15 (18.1) 0.79 (0.37–1.67) 15 (22.7) 1.35 (0.63–2.88) 0.08 TXB2 63.33 51 (32.7) 50 (31.8) 1.00 (Reference) 24 (28.9) 1.00 (Reference) 22 (33.3) 1.00 (Reference) >63.33–1,350 52 (33.3) 54 (34.4) 1.02 (0.56–1.84) 31 (37.4) 1.41 (0.68–2.93) 20 (30.3) 0.70 (0.32–1.53) >1350 53 (34.0) 53 (33.8) 1.08 (0.60–1.94) 28 (33.7) 1.28 (0.61–2.68) 24 (36.4) 1.04 (0.49–2.20) 0.59 Ptrend 0.81 0.49 0.94 Arachidonic acid/LOX derived 15-HETE 300.17 52 (33.3) 59 (37.6) 1.00 (Reference) 29 (34.9) 1.00 (Reference) 28 (42.4) 1.00 (Reference) >300.17–448.33 51 (32.7) 35 (22.3) 0.62 (0.34–1.12) 20 (24.1) 0.75 (0.36–1.58) 11 (16.7) 0.43 (0.19–0.98) >448.33 53 (34.0) 63 (40.1) 1.18 (0.67–2.08) 34 (41.0) 1.31 (0.65–2.63) 27 (40.9) 1.08 (0.53–2.17) 0.63 Ptrend 0.55 0.50 0.88 12-HETE 2,888.33 52 (33.3) 56 (35.7) 1.00 (Reference) 28 (33.7) 1.00 (Reference) 24 (36.4) 1.00 (Reference) >2,888.33–7,700 51 (32.7) 42 (26.8) 0.75 (0.41–1.35) 21 (25.3) 0.78 (0.37–1.64) 20 (30.3) 0.80 (0.38–1.70) >7,700 53 (34.0) 59 (37.6) 1.15 (0.65–2.02) 34 (41.0) 1.39 (0.69–2.79) 22 (33.3) 0.98 (0.47–2.04) 0.37 Ptrend 0.64 0.42 0.91 8-HETEf 3,581.67 52 (33.3) 43 (27.4) 1.00 (Reference) 18 (21.7) 1.00 (Reference) 24 (36.4) 1.00 (Reference) >3,581.67–6,016.67 51 (32.7) 46 (29.3) 1.36 (0.74–2.49) 28 (33.7) 2.60 (1.18–5.73) 17 (25.8) 0.75 (0.34–1.63) >6,016.67 53 (34.0) 68 (43.3) 1.82 (1.01–3.26) 37 (44.6) 2.53 (1.18–5.39) 25 (37.9) 1.15 (0.56–2.36) 0.10 Ptrend 0.04 0.03 0.75 5-HETE 498.33 52 (33.3) 36 (22.9) 1.00 (Reference) 19 (22.9) 1.00 (Reference) 15 (22.7) 1.00 (Reference) >498.33–780.00 51 (32.7) 69 (43.9) 2.25 (1.24–4.07) 35 (42.2) 2.22 (1.06–4.65) 30 (45.5) 2.45 (1.13–5.31) >780.00 53 (34.0) 52 (33.1) 1.59 (0.86–2.94) 29 (34.9) 1.72 (0.80–3.69) 21 (31.8) 1.60 (0.71–3.62) 0.93 Ptrend 0.18 0.16 0.27 LTB-4 78.83 52 (33.3) 40 (25.5) 1.00 (Reference) 20 (24.1) 1.00 (Reference) 16 (24.2) 1.00 (Reference) >78.83–236.17 51 (32.7) 56 (35.7) 1.67 (0.91–3.07) 27 (32.5) 1.73 (0.80–3.76) 26 (39.4) 1.90 (0.87–4.15) >236.17 53 (34.0) 61 (38.9) 1.68 (0.93–3.04) 36 (43.4) 2.23 (1.07–4.67) 24 (36.4) 1.59 (0.73–3.47) 0.75 Ptrend 0.10 0.03 0.22 Arachidonic acid/P450 derived 5,6-DHET 268.17 52 (33.3) 44 (28.0) 1.00 (Reference) 18 (21.7) 1.00 (Reference) 26 (39.4) 1.00 (Reference) >268.17–416.50 51 (32.7) 56 (35.7) 1.33 (0.74–2.42) 36 (43.3) 2.48 (1.16–5.33) 19 (28.8) 0.73 (0.34–1.56) >416.50 53 (34.0) 57 (36.3) 1.43 (0.79–2.62) 29 (34.9) 1.85 (0.84–4.08) 21 (31.8) 0.93 (0.44–1.94) 0.05 Ptrend 0.25 0.19 0.98 8,9-EET 526.67 52 (33.3) 54 (34.4) 1.00 (Reference) 23 (27.7) 1.00 (Reference) 28 (42.4) 1.00 (Reference) >526.67–820 51 (32.7) 42 (26.8) 0.77 (0.43–1.41) 24 (28.9) 1.08 (0.51–2.29) 16 (24.2) 0.63 (0.27–1.27) >820 53 (34.0) 61 (38.9) 1.21 (0.67–2.17) 36 (43.4) 1.59 (0.77–3.27) 22 (33.3) 0.90 (0.43–1.87) 0.11 Ptrend 0.52 0.22 0.90 8,9-DHET 186.5 52 (33.3) 42 (26.8) 1.00 (Reference) 20 (24.1) 1.00 (Reference) 20 (30.3) 1.00 (Reference) >186.5–264.5 51 (33.3) 61 (38.9) 1.65 (0.91–3.00) 34 (41.0) 1.84 (0.86–3.91) 26 (39.4) 1.62 (0.77–3.42) >264.5 53 (34.0) 54 (34.4) 1.46 (0.80–2.68) 29 (34.9) 1.57 (0.72–3.40) 20 (30.3) 1.19 (0.55–2.59) 0.32 Ptrend 0.25 0.32 0.56 (Continued on the following page)

194 Cancer Epidemiol Biomarkers Prev; 28(1) January 2019 Cancer Epidemiology, Biomarkers & Prevention

Serum Fatty Acid Metabolites and Ovarian Cancer Risk

Table 3. Association of arachidonic acid and linoleic acid–derived metabolites and risk of overall ovarian cancer and histologic subtypes (serous and nonserous) of ovarian cancer in a nested case–control study in the PLCO Cancer Screening Trial (Cont'd ) Control Cases Overall cases Serous cases Nonserous cases Metabolites (pg/ml) N (%) N (%) ORa (95% CI)a N (%) OR (95% CI) N (%) OR (95% CI) P-Hetb 11,12-EET 401.67 52 (33.3) 47 (29.9) 1.00 (Reference) 25 (30.1) 1.00 (Reference) 19 (28.8) 1.00 (Reference) >401.67–805 51 (32.7) 56 (35.7) 1.20 (0.67–2.14) 27 (32.5) 1.09 (0.53–2.25) 28 (42.4) 1.49 (0.71–3.10) >805 53 (34.0) 54 (34.4) 1.13 (0.62–2.04) 31 (37.4) 1.15 (0.56–2.35) 19 (28.8) 0.98 (0.45–2.15) 0.61 Ptrend 0.70 0.68 0.87 11,12-DHET 334.33 52 (33.3) 52 (33.1) 1.00 (Reference) 25 (30.1) 1.00 (Reference) 25 (37.9) 1.00 (Reference) >334.33–441.67 51 (32.7) 52 (33.1) 1.12 (0.63–2.01) 29 (34.9) 1.29 (0.63–2.65) 22 (33.3) 0.97 (0.47–2.01) >441.67 53 (34.0) 53 (33.8) 1.16 (0.65–2.10) 29 (34.9) 1.35 (0.65–2.80) 19 (28.8) 0.82 (0.39–1.76) 0.06 Ptrend 0.61 0.55 0.74 14,15-EET 590 52 (33.3) 54 (34.4) 1.00 (Reference) 25 (30.1) 1.00 (Reference) 25 (37.9) 1.00 (Reference) >590–1,765 51 (32.7) 43 (27.4) 0.74 (0.41–1.34) 23 (27.7) 0.84 (0.39–1.77) 19 (28.8) 0.78 (0.37–1.66) >1765 53 (34.0) 60 (38.2) 1.01 (0.56–1.81) 35 (42.2) 1.13 (0.55–2.33) 22 (33.3) 0.85 (0.40–1.80) 0.38 Ptrend 0.99 0.74 0.71 14,15-DHET 448.33 53 (34.0) 52 (33.1) 1.00 (Reference) 24 (28.9) 1.00 (Reference) 25 (37.9) 1.00 (Reference) >448.33–578.33 50 (32.1) 51 (32.5) 1.01 (0.56–1.81) 29 (34.9) 1.03 (0.49–2.17) 21 (31.8) 0.98 (0.47–2.06) >578.33 53 (34.0) 54 (34.4) 1.26 (0.70–2.27) 30 (36.1) 1.60 (0.77–3.35) 20 (30.3) 0.88 (0.41–1.88) 0.05 Ptrend 0.45 0.26 0.87 Arachidonic acid/nonenzymatic derived 11-HETE 136.50 52 (33.3) 52 (33.1) 1.00 (Reference) 27 (32.5) 1.00 (Reference) 24 (36.4) 1.00 (Reference) >136.50–244.67 51 (32.7) 48 (30.6) 1.11 (0.62–2.00) 23 (27.7) 1.06 (0.51–2.19) 19 (28.8) 0.94 (0.42–1.99) >244.67 53 (34.0) 57 (36.3) 1.23 (0.69–2.20) 33 (39.8) 1.41 (0.69–2.89) 23 (34.9) 1.10 (0.52–2.30) 0.85 Ptrend 0.49 0.32 0.91 c 8-iso-PGF2a 0 145 (92.9) 147 (93.6) 1.00 (Reference) 76 (91.6) 1.00 (Reference) 63 (95.5) 1.00 (Reference) >0 (detectable) 11 (7.1) 10 (6.4) 0.81 (0.31–2.10) 7 (8.4) 1.40 (0.47–4.15) 3 (4.6) 0.34 (0.07–1.66) 0.27 aORs and 95% CIs were estimated using unconditional logistic regression model adjusted for matching factors [age at blood collection (55–59, 60–64, 65–69, 70þ years), race (white, black, and other), study center, and time (a.m., p.m.) and date (3-month categories of blood collection)] and a priori selected risk factors (parity, duration of oral contraceptive use, duration of menopausal hormone therapy use, cigarette smoking, and BMI). bStatistical test for heterogeneity across histologic subtypes was based on the Wald test. cMetabolites with 90% or more individuals with nondetectable levels were categorized into detectable versus undetectable. dTertile categories of the metabolites were based on the distribution among controls. eP-values for trend across tertile as continuous variable were calculated using the Wald test. fFDR q-values: 8-HETE ¼ 0.17; 12,13-DHOME, 13-HODE, and 9,12,13-THOME ¼ 0.06; and 9-HODE ¼ 0.15. increased ovarian cancer risk with the linoleic acid pathways Linoleic acid oxidation by cytochrome P450 monoxygenase based on meta-analysis. produces 12,13-DHOME (isoleukotoxin diol). A study of ovarian Free fatty acid metabolism by the LOX pathway produces 8- tumors (n ¼ 167 cases) reported overexpression of the cyto- HETE, 13-HODE, 9-HODE, and 9,12,13-THOME. Although the chrome P450 allelic variant (CYP1B1; ref. 27). Several studies role of the LOX pathway in ovarian cancer has not been exten- have demonstrated the role of 12,13-DHOME in suppression of sively evaluated, few studies support a pivotal role of the LOX the immune system, increased adipocyte differentiation, cell pathway in ovarian cancer prognosis. Speciﬁcally, a tissue micro- proliferation, and apoptosis (28, 29), and as a ligand for array study of 245 parafﬁn-embedded epithelial ovarian cancer PPAR-g (30). The role of 12,13-DHOME in several cellular func- samples demonstrated that strong expression of LOX receptors tions might contribute to the increased risk of ovarian cancer. was indicative of worse ovarian cancer prognosis (23). An immu- Furthermore, the positive association of 12,13-DHOME with risk nohistochemistry (IHC) analysis comparing expression of 12- of ovarian cancer may be related to increased production of 12,13- LOX in a serous ovarian cancer cell line and normal ovarian DHOME by CYP1B1. epithelium demonstrated that expression of 12-LOX was higher in To our knowledge no prior study has explored the association serous ovarian carcinoma compared to normal ovarian epitheli- between pre-diagnostic levels of circulating fatty acid metabolites um and regulated the cell growth through ERK and MAPK sig- from pathways involving COX, LOX, and cytochrome P450 naling (24). The LOX pathway metabolites (8-HETE,13-HODE, 9- enzymes and ovarian cancer risk. The strengths of our study HODE, and 9,12,13-THOME) are also known as PPAR-a/g include a well-designed study nested within the large prospective ligands, a lipid-activated transcription factor. PPAR-a promotes PLCO study, with detailed information on established ovarian lipid uptake (25), whereas PPAR-g is mostly implicated in lipid cancer risk factors enabling careful control for confounding and storage and adipocyte differentiation (26). Experimental data multiple comparisons. All the analytes were measured by sensitive suggest that PPAR isotypes have a role in tumor suppression tandem mass spectrometry. We matched cases and controls on and/or progression, but the mechanism is unclear (22–24). time of day and date of blood collection to minimize any effects Therefore, the metabolites produced by the LOX pathway may due to variability in storage conditions. To limit the impact of have a role in ovarian carcinogenesis via PPAR-a/g activation. reverse causation, we excluded cases diagnosed within 2 years of

www.aacrjournals.org Cancer Epidemiol Biomarkers Prev; 28(1) January 2019 195

Hada et al.

Figure 2. Association of arachidonic acid and linoleic acid pathways and risk of ovarian cancer overall using ﬁxed effects meta-analysis in a nested case-control study in the PLCO Cancer Screening Trial. A, P-values for the summary odds ratio are reported. B, ORs and 95% CIs for a one tertile increase in metabolite level were estimated using unconditional logistic regression adjusted for matching factors [age at blood collection (55–59, 60–64, 65–69, 70þ years), race (white, black, and other), study center, and time (a.m., p.m.) and date (3-month categories) of blood collection] and a priori selected risk factors (parity, duration of oral contraceptive use, duration of menopausal hormone therapy use, and cigarette smoking). , ORs and 95% CIs for 17,18-

EpETE, PGD2, and 8-iso-PGF2a compare detectable versus nondetectable metabolite levels.

blood draw. Our study also included limitations. Although we lating levels of arachidonic acid (31). Thus, validation in an included all available cases, the number of cases included in independent cohort is needed. the current study was limited. There is a possibility that the In conclusion, our data provide evidence that out of 31 fatty single time point measurement of these metabolites may not acid metabolites studied, five metabolites (8-HETE, 12,13- reflect their concentration over a longer period. To our knowl- DHOME, 13-HODE, 9-HODE, and 9,12,13-THOME), generated edge, there are no studies on temporal variability of serum via the LOX/cytochrome P450 linoleic acid metabolite pathway, concentrations of metabolites within an individual but a were associated with the increased risk of ovarian cancer. In systematic review looking at the change in tissue arachidonic pathway analysis, ovarian cancer risk was associated with all three acid after the consumption of western diet showed that linoleic acid pathways (COX derived, LOX derived, and cyto- decreasing the intake of linoleic acid had no effect on circu- chrome P450 derived). As inflammation is a well-established risk

196 Cancer Epidemiol Biomarkers Prev; 28(1) January 2019 Cancer Epidemiology, Biomarkers & Prevention

Serum Fatty Acid Metabolites and Ovarian Cancer Risk

factor for ovarian cancer, these precursors of inflammatory mar- Writing, review, and/or revision of the manuscript: M. Hada, M.L. Edin, kers associated with ovarian cancer might be important to under- P. Hartge, N. Wentzensen, D.C. Zeldin, B. Trabert stand a potential mechanism of the etiology of ovarian cancer. Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.L. Edin, B. Trabert Study supervision: D.C. Zeldin, B. Trabert Disclosure of Potential Conflicts of Interest No potential conflicts of interest were disclosed. Acknowledgments This work was supported by the Intramural Research Program of the National Authors' Contributions Cancer Institute. The authors would like to thank the NIH Fellows Editorial Conception and design: P. Hartge, B. Trabert Board for comments on a draft of the authors' manuscript. Development of methodology: M. Hada, M.L. Edin, B. Trabert Acquisition of data (provided animals, acquired and managed patients, The costs of publication of this article were defrayed in part by the payment of provided facilities, etc.): M.L. Edin, P. Hartge, F.B. Lih, N. Wentzensen, page charges. This article must therefore be hereby marked advertisement in D.C. Zeldin, B. Trabert accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Analysis and interpretation of data (e.g., statistical analysis, biostatis- tics, computational analysis): M. Hada, M.L. Edin, N. Wentzensen, Received May 11, 2018; revised August 10, 2018; accepted September 19, B. Trabert 2018; published first September 27, 2018.

References 1. Coussens LM, Werb Z. Inflammation and cancer. Nature 2002;420:860–7. impairs cardiac recovery after ischemia. J Biol Chem 2018;293: 2. Crusz SM, Balkwill FR. Inflammation and cancer: advances and new agents. 3281–92. Nat Rev Clin Oncol 2015;12:584–96. 19. Jerschow E, Edin ML, Pelletier T, Abuzeid WM, Akbar NA, Gibber M, et al. 3. Kisielewski R, Tolwinska A, Mazurek A, Laudanski P. Inflammation and Plasma 15-Hydroxyeicosatetraenoic acid predicts treatment outcomes in ovarian cancer—current views. Ginekol Pol 2013;84:293–7. aspirin-exacerbated respiratory disease. J Allergy Clin Immunol Prac- 4. Maccio A, Madeddu C. Inflammation and ovarian cancer. Cytokine 2012; tice;5:998–1007.e2. 58:133–47. 20. Zha W, Edin ML, Vendrov KC, Schuck RN, Lih FB, Jat JL, et al. Functional 5. Barry JA, Azizia MM, Hardiman PJ. Risk of endometrial, ovarian and breast characterization of cytochrome P450-derived epoxyeicosatrienoic acids in cancer in women with polycystic ovary syndrome: a systematic review and adipogenesis and obesity. J Lipid Res 2014;55:2124–36. meta-analysis. Hum Reprod Update 2014;20:748–58. 21. Edin ML, Wang Z, Bradbury JA, Graves JP, Lih FB, DeGraff LM, et al. 6. Trabert B, Ness RB, Lo-Ciganic WH, Murphy MA, Goode EL, Poole EM, et al. Endothelial expression of human cytochrome P450 epoxygenase CYP2C8 Aspirin, nonaspirin nonsteroidal anti-inflammatory drug, and acetamin- increases susceptibility to ischemia-reperfusion injury in isolated mouse ophen use and risk of invasive epithelial ovarian cancer: a pooled analysis heart. FASEB J 2011;25:3436–47. in the Ovarian Cancer Association Consortium. J Natl Cancer Inst 22. Trabert B, Pinto L, Hartge P, Kemp T, Black A, Sherman ME, et al. Pre- 2014;106:djt431. diagnostic serum levels of inflammation markers and risk of ovarian cancer 7. Ricciotti E, FitzGerald GA. Prostaglandins and inflammation. Arterioscler in the prostate, lung, colorectal and ovarian cancer (PLCO) screening trial. Thromb Vasc Biol 2011;31:986–1000. Gynecol Oncol 2014;135:297–304. 8. Wang D, Dubois RN. Eicosanoids and cancer. Nat Rev Cancer 2010; 23. Rocconi RP, Kirby TO, Seitz RS, Beck R, Straughn JM Jr, Alvarez RD, et al. 10:181–93. Lipoxygenase pathway receptor expression in ovarian cancer. Reprod Sci 9. Hyde CA, Missailidis S. Inhibition of arachidonic acid metabolism and its 2008;15:321–6. implication on cell proliferation and tumour-angiogenesis. Int Immuno- 24. Guo AM, Liu X, Al-Wahab Z, Maddippati KR, Ali-Fehmi R, Scicli AG, et al. pharmacol 2009;9:701–15. Role of 12-lipoxygenase in regulation of ovarian cancer cell proliferation 10. Di Pasquale MG. The essentials of essential fatty acids. J Diet Suppl 2009; and survival. Cancer Chemother Pharmacol 2011;68:1273–83. 6:143–61. 25. Staels B, Dallongeville J, Auwerx J, Schoonjans K, Leitersdorf E, Fruchart JC. 11. Calder PC. Omega-3 fatty acids and inflammatory processes. Nutrients Mechanism of action of fibrates on lipid and lipoprotein metabolism. 2010;2:355–74. Circulation 1998;98:2088–93. 12. Simopoulos AP. Omega-3 fatty acids in inflammation and autoimmune 26. Rangwala SM, Lazar MA. Peroxisome proliferator-activated receptor diseases. J Am Coll Nutr 2002;21:495–505. gamma in diabetes and metabolism. Trends Pharmacol Sci 2004;25: 13. Harizi H, Corcuff JB, Gualde N. Arachidonic-acid-derived eicosanoids: 331–6. roles in biology and immunopathology. Trends Mol Med 2008;14:461–9. 27. McFadyen MC, Cruickshank ME, Miller ID, McLeod HL, Melvin WT, Haites 14. Greene ER, Huang S, Serhan CN, Panigrahy D. Regulation of inflammation NE, et al. Cytochrome P450 CYP1B1 over-expression in primary and in cancer by eicosanoids. Prostaglandins Other Lipid Mediators 2011; metastatic ovarian cancer. Br J Cancer 2001;85:242–6. 96:27–36. 28. Lynes MD, Leiria LO, Lundh M, Bartelt A, Shamsi F, Huang TL, et al. The 15. Prorok PC, Andriole GL, Bresalier RS, Buys SS, Chia D, Crawford ED, et al. cold-induced lipokine 12,13-diHOME promotes fatty acid transport into Design of the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer brown adipose tissue. Nat Med 2017;23:631–7. Screening Trial. Control Clin Trials 2000;21:273S–309S. 29. Moran JH, Weise R, Schnellmann RG, Freeman JP, Grant DF. Cytotoxicity 16. Hayes RB, Sigurdson A, Moore L, Peters U, Huang WY, Pinsky P, et al. of linoleic acid diols to renal proximal tubular cells. Toxicol Appl Phar- Methods for etiologic and early marker investigations in the PLCO trial. macol 1997;146:53–9. Mutat Res 2005;592:147–54. 30. Medina-Gomez G, Gray S, Vidal-Puig A. Adipogenesis and lipotoxicity: 17. Rauzi F, Kirkby NS, Edin ML, Whiteford J, Zeldin DC, Mitchell JA, et al. role of peroxisome proliferator-activated receptor g (PPARg) and Aspirin inhibits the production of proangiogenic 15(S)-HETE by platelet PPARgcoactivator-1 (PGC1). Public Health Nutr 2007;10:1132–7. cyclooxygenase-1. FASEB J 2016;30:4256–66. 31. Rett BS, Whelan J. Increasing dietary linoleic acid does not increase tissue 18. Edin ML, Hamedani BG, Gruzdev A, Graves JP, Lih FB, Arbes SJ, et al. arachidonic acid content in adults consuming Western-type diets: a sys- Epoxide hydrolase 1 (EPHX1) hydrolyzes epoxyeicosanoids and tematic review. Nutr Metab (Lond) 2011;8:36.

www.aacrjournals.org Cancer Epidemiol Biomarkers Prev; 28(1) January 2019 197

Prediagnostic Serum Levels of Fatty Acid Metabolites and Risk of Ovarian Cancer in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial

Manila Hada, Matthew L. Edin, Patricia Hartge, et al.

Cancer Epidemiol Biomarkers Prev 2019;28:189-197. Published OnlineFirst September 27, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/1055-9965.EPI-18-0392

Supplementary Access the most recent supplemental material at: Material http://cebp.aacrjournals.org/content/suppl/2018/09/27/1055-9965.EPI-18-0392.DC1

Cited articles This article cites 30 articles, 4 of which you can access for free at: http://cebp.aacrjournals.org/content/28/1/189.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://cebp.aacrjournals.org/content/28/1/189.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department Subscriptions at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cebp.aacrjournals.org/content/28/1/189. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.