Premium
Association Between the Serum Metabolome and All‐Cause Mortality: A Prospective Analysis in the Alpha‐Tocopherol, Beta‐Carotene Cancer Prevention (ATBC) Study Cohort
Author(s) -
Huang Jiaqi,
Weinstein Stephanie J.,
Moore Steven C.,
Sampson Joshua N.,
Albanes Demetrius
Publication year - 2017
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.31.1_supplement.941.2
Subject(s) - medicine , cancer prevention , metabolomics , metabolite , prospective cohort study , proportional hazards model , cancer , confounding , oncology , bioinformatics , biology
Tobacco smoking, hypertension, hyperglycemia, overweight, and inactivity are the leading causes of mortality worldwide. Beyond these factors, prospective metabolomic profiling may help identify individuals at higher risk of death and point to molecular factors and biological pathways that have preventive potential. We conducted a serum metabolomic analysis of 667 men included as controls in 6 nested case‐control studies in the Alpha‐Tocopherol, Beta‐Carotene Cancer Prevention (ATBC) Study. During up to 28 years of follow‐up, there were 457 deaths, including 199 deaths from cardiovascular disease (CVD) and 112 deaths from cancer. Ultra‐high performance LC/MS and GC/MS (Metabolon, Inc., NC) identified 1,173 known metabolites, and 375 metabolites were retained for analysis after excluding those missing in ≥2 of the study sets. We modeled the relationship between each metabolite and risk of death by cox regression and report the set of metabolites where the False Discovery Rate (FDR) q‐value<0.02. We next split the data into training and test sets. Using only the training set, we selected those metabolites (N=14) with an FDR q‐value of <0.15 and included all selected metabolites in a multivariable cox regression. In the test set, we constructed a metabolomic score, where each individual's metabolite levels were weighted by the coefficients from the regression model, and categorized individuals into tertiles according to their score. All models were adjusted for mortality risk factors. 7‐Methylguanine, N‐acetylvaline, taurocholate, N‐acetyltryptophan, 5,6‐dihydrothymine, homocitrulline, dimethylglycine, myristoleate (14:1n5), palmitoleate (16:1n7), C‐glycosyltryptophan, taurochenodeoxycholate, 5‐dodecenoate (12:1n7), erythronate, N1‐methylguanosine, N‐formylmethionine, and hexanoylcarnitine yielded the strongest metabolite risk signals at a threshold of FDR q‐value <0.02 (0.05≤β≤0.07, and 2.0×10 −6 ≤ P ≤4.6×10 −4 , respectively). The validation analysis in the test‐set showed that individuals in the 2 nd and 3 rd tertiles of the metabolite risk score experienced twice the mortality of those in the lowest tertile (2 nd tertile score: HR=2.0, 95% CI: 1.2–3.2; 3 rd tertile highest score: HR=2.2, 95% CI: 1.3–3.6). Our analysis identified several serum metabolites independently associated with overall mortality, as was their composite risk score, which replicated in the independent test‐set. The metabolite signals are associated with pathways related to CVD, physical activity, ageing, inflammation and nucleoside tumor markers. Research support: Intramural Research Program of the National Cancer Institute, NIH