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

Study design

We conducted a nested case–control study within the screening arm of the PLCO Cancer Screening Trial, a randomized two-arm screening trial of men and women, ranging in age 55 to 74 years was conducted between 1993 and 2001 (15). Briefly, 78,216 women recruited from 10 centers across the United States (Birmingham, AL; Denver, CO; Detroit, MI; Honolulu, HI; Marshfield, WI; Minneapolis, MN; Pittsburgh, PA; Salt Lake City, UT; St. Louis, MO; and Washington, DC). Participants in the screening arm provided a blood sample at baseline and during five follow-up medical examinations and were stored at −70°C (16). All participants completed a self-administered baseline questionnaire. Cancer cases were identified by annual mailed questionnaires which were then confirmed by linking to population-based registries and the National Death Index. Medical and pathologic records were obtained when possible.

To avoid the possibility of reverse causality, cases for our study were 157 individuals diagnosed between 2 and 14 years after blood collection with follow-up through December 2010. To ensure a relatively equal distribution of specimens between 2 and 14 years prior to diagnosis, 10.9% of samples selected were measured at baseline and the remaining at follow-up visits (16.3% year 1, 25.9% year 2, 13.7% year 3, and 33.2% year 4). For inclusion in our study, serum samples were selected from a prediagnostic blood draw given the availability of unthawed serum, consent to biochemical studies, completion of the baseline questionnaire, and no history of cancer (except nonmelanoma skin cancer). There were 160 cases identified out of which three were excluded from the study because the sample volume was not sufficient. Controls were women with no history of oophorectomy at the time of diagnosis of their matched case and were matched on age at blood collection (55–59, 60–64, 65–69, 70+ years), race (white, black, other), study center, and time (a.m. and p.m.) and date (3-month categories) of blood collection. Serum specimens from a single visit were measured for each study subject. All the study participants provided written informed consent and the study received approval from the institutional review board of the involved study centers and from the National Cancer Institute.

Laboratory analysis

Levels of 34 free fatty acid metabolites generated through three different pathways—COX pathway, cytochrome P450 pathway, and LOX pathway—were measured in serum. All free fatty acid metabolites were separated by electrospray ionization LC (Agilent 1200 series capillary HPLC) and quantified using tandem mass spectrometry (MS/MS, MDS Sciex API 3000, negative ion mode) with a Turbolon Spray and scheduled multiple reaction. The details of the assay have been published previously (17–19). Concentrations (pg/mL) were quantified with Analyst software (v 1.5; Applied Biosystem) using metabolite and internal standard peaks for each sample.

To evaluate assay performance, we included duplicate quality control (QC) samples in each batch and calculated the within-batch and between-batch coefficients of variation (CV). We also calculated the intraclass correlation coefficient (ICC) for each analyte. 51.6% of CVs were <10%, 25.8% were 10% to 15%, and 22.5% were 15% to 20% (Supplementary Table S1). We excluded from evaluation metabolite 5,6-EET given known limitations in its detection via LC/MS-MS, we further excluded two metabolites (19-HETE and 20-HETE) that were undetectable in nearly all samples. Thus, after exclusions, 31 metabolites were evaluated (17, 20, 21).

Statistical analysis

We used generalized linear models adjusted for age to calculate geometric mean concentrations of free fatty acid metabolites by case–control status and by other factors (e.g., smokers vs. nonsmokers, normal BMI versus overweight/obese BMI, and aspirin/Ibuprofen users vs. nonusers). P-values were calculated using a Wald test.

We used unconditional logistic regression models to calculate odds ratios and 95% confidence intervals (CI) for the association between prediagnostic serum levels of free fatty acid metabolites and risk of ovarian cancer. All models were adjusted for matching factors and potential confounders: parity (nulliparous/parous), duration of oral contraceptive use (never, 1–5 years, 6+ years), duration of menopausal hormone therapy use (never, 1–5 years, 6+ years), cigarette smoking status (never, former, current), and body mass index (BMI; <25, 25–29.9, 30+ kg/m²). The risk of ovarian cancer was examined across tertile categories of the metabolites based on the distribution among controls. Metabolites with 90% or more individuals with nondetectable levels were categorized into two groups (detectable vs. undetectable). A Wald test using tertile categories as a continuous variable was used to evaluate trend. For markers associated with ovarian cancer risk, we further adjusted for the inflammatory markers (CRP, TNFα, and IL8) that were positively associated with ovarian cancer risk in a previous analysis (22).

We also explored the association between free fatty acid metabolites and risk of ovarian cancer by histologic subtypes (serous vs. nonserous) and number of years from blood collection to cancer diagnosis (cancer diagnosis 2–5 years vs. >5 years from blood draw). P-values for heterogeneity were based on the Wald test for the ordinal metabolite variable from a case-only model with serous histology or the shortest time (2–5 years) between blood draw and diagnosis as the reference group.

We used fixed effect meta-analysis to calculate summary objective responses (OR) of the association of fatty acid metabolite pathways (Fig. 1) with the risk of ovarian cancer overall and by histologic subtypes.

We evaluated effect modification by BMI (normal vs. overweight/obese) and use of medication (aspirin/ibuprofen users vs. aspirin/ibuprofen nonusers) using stratified models. Likelihood ratio tests were calculated to test for interaction using the cross-product term. Because of a limited number of current smokers, we could not evaluate effect modification by smoking status, but we conducted a sensitivity analysis restricted to never/former smokers. As a secondary analysis, we evaluated the association between free fatty acid metabolites and the risk of ovarian cancer adjusting for aspirin use.

To account for multiple comparisons, we applied the false discovery rate (FDR) for the primary free fatty acid metabolites associations with the ovarian cancer. All other analyses were considered exploratory and not corrected for multiple comparisons. All tests were two-sided and statistical significance was defined using a P-value <0.05. Analyses were conducted in SAS version 9.4 (SAS Inc.).