Comparison of the Predictive Value of GlycA and ... - Clinical Chemistry

0 downloads 0 Views 636KB Size Report
12 May 2016 - Comparison of the Predictive Value of GlycA and. Other Biomarkers of ... ChrIRD, and total cancer after adjustment for hsCRP,. IL-6, and ...... Research Funding: NMR GlycA values provided at no cost by Lipo- science Inc (now ...
Papers in Press. Published May 12, 2016 as doi:10.1373/clinchem.2016.255828 The latest version is at http://hwmaint.clinchem.org/cgi/doi/10.1373/clinchem.2016.255828 Clinical Chemistry 62:7 000 – 000 (2016)

Other Areas of Clinical Chemistry

Comparison of the Predictive Value of GlycA and Other Biomarkers of Inflammation for Total Death, Incident Cardiovascular Events, Noncardiovascular and Noncancer Inflammatory-Related Events, and Total Cancer Events Daniel A Duprez,1* James Otvos,2 Otto A. Sanchez,3 Rachel H. Mackey,4 Russell Tracy,5 and David R. Jacobs Jr3

BACKGROUND: GlycA is a biomarker that reflects integrated concentrations and glycosylation states of several acute-phase proteins. We studied the association of GlycA and inflammatory biomarkers with future death and disease. METHODS: A total of 6523 men and women in the Multi-Ethnic Study of Atherosclerosis who were free of overt cardiovascular disease (CVD) and in generally good health had a baseline blood sample taken. We assayed high-sensitivity C-reactive protein (hsCRP), interleukin-6 (IL-6), and D-dimer. A spectral deconvolution algorithm was used to quantify GlycA signal amplitudes from automated nuclear magnetic resonance (NMR) LipoProfile® test spectra. Median follow-up was 12.1 years. Among 4 primary outcomes, CVD events were adjudicated, death was by death certificate, and chronic inflammatory-related severe hospitalization and death (ChrIRD) and total cancer were classified using International Classification of Diseases (ICD) codes. We used Poisson regression to study baseline GlycA, hsCRP, IL-6, and D-dimer in relation to total death, CVD, ChrIRD, and total cancer. RESULTS:

Relative risk per SD of GlycA, IL-6, and D-dimer for total death (n ⫽ 915); for total CVD (n ⫽ 922); and for ChrIRD (n ⫽ 1324) ranged from 1.05 to 1.20, independently of covariates. In contrast, prediction from hsCRP was statistically explained by adjustment for other inflammatory variables. Only GlycA was predictive for total cancer (n ⫽ 663). Women had 7% higher values of all inflammatory

1

Cardiovascular Division, School of Medicine, University of Minnesota, MN; 2 LabCorp, Raleigh, NC; 3 Department of Internal Medicine, Division of Nephrology, School of Medicine University of Minnesota, MN; 4 University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA; 5 Department of Pathology & Laboratory Medicine, and Biochemistry, University of Vermont College of Medicine, Colchester, VT. * Address correspondence to this author at: Cardiovascular Division, Medical School, University of Minnesota, 420 Delaware St SE, MMC 508, Minneapolis, MN 55455. Fax +612-626-4411, e-mail [email protected].

biomarkers than men and had a significantly lower GlycA prediction coefficient than men in predicting total cancer. CONCLUSIONS: The composite biomarker GlycA derived from NMR is associated with risk for total death, CVD, ChrIRD, and total cancer after adjustment for hsCRP, IL-6, and D-dimer. IL-6 and D-dimer contribute information independently of GlycA.

© 2016 American Association for Clinical Chemistry

Inflammation is a common mechanism in infections and many chronic diseases, including chronic respiratory disease, gastrointestinal disease, autoimmune disease, cancer, and cardiovascular disease (CVD)6 (1– 4 ). Epidemiological studies have shown the association of systemic inflammation and adverse clinical outcome by demonstrating consistent independent associations of highsensitivity C-reactive protein (hsCRP), interleukin (IL)-6, and D-dimer with both incident CVD and overall mortality (5–7 ). Nuclear magnetic resonance (NMR) spectroscopy can contribute to the study of inflammation. NMR has proven to be an efficient research tool for quantifying numerous metabolites in blood samples. Up to now, the only application in routine clinical use has been lipoprotein particle analysis by NMR LipoProfile® testing (8 ). Otvos et al. (9 ) recently reported the quantification of a specific NMR signal, named GlycA, which occurs in the NMR spectrum used to assess lipoprotein particles. The resonance in plasma NMR spectra identified as GlycA

Received February 5, 2016; accepted April 12, 2016. Previously published online at DOI: 10.1373/clinchem.2016.255828 © 2016 American Association for Clinical Chemistry 6 Nonstandard abbreviations: CVD, cardiovascular disease; hsCRP, C-reactive protein; IL, interleukin; NMR, nuclear magnetic resonance; MESA, Multi-Ethnic Study of Atherosclerosis; ChrIRD, chronic inflammatory-related severe hospitalization and death; ICD, International Classification of Diseases; CHD, coronary heart disease; IPP, improvement in prediction probability; WHS, Women’s Health Study.

1

Copyright (C) 2016 by The American Association for Clinical Chemistry

originates from the N-acetyl methyl group protons of mobile glycan residues of glycoproteins (10, 11 ). GlycA is a composite biomarker that senses the integrated concentrations and glycosylation states of several of the most abundant acute-phase proteins in serum. The limited available information about GlycA finds that it is associated with incident CVD and risk for type 2 diabetes mellitus in initially healthy women (12–14 ). Using data from the Multi-Ethnic Study of Atherosclerosis (MESA) (15 ) in men, women, and 4 racial/ ethnic groups, we aimed to examine the association of baseline GlycA concentration with total death, incident CVD, chronic inflammatory-related non-CVD, and noncancer disease chronic inflammatory-related severe hospitalization and death (ChrIRD) events, and total cancer and to compare these associations with other commonly used clinical biomarkers of chronic inflammation (hsCRP, IL-6, and D-dimer). In addition, we aimed to evaluate the associations of each of the 4 markers with the risk of future outcomes adjusting for the other 3 biomarkers to assess which of them provide additional clinical utility beyond the information conveyed by the other inflammatory biomarkers. We hypothesized that GlycA predicts each outcome independently of the other 3 biomarkers. Material and Methods STUDY SAMPLE

MESA was initiated to investigate the prevalence, correlates, and progression of subclinical CVD in people initially free of overt clinical CVD, although they may have had non-CVD conditions before baseline (15 ). Between 2000 and 2002, 6814 men and women of white, black, Hispanic, or Chinese race/ethnicity, aged 45– 84 years, were recruited from 6 US communities. Participants missing any GlycA variable or any covariates or without any follow-up were excluded, leaving n ⫽ 6523 (the largest contribution to exclusions was missing IL-6). The institutional review boards at all participating centers approved the study, and all participants gave informed consent. LABORATORY MEASUREMENTS

Lipids, lipoprotein particles, and other laboratory assays. Blood was drawn after a 12-h fast, and EDTA plasma was stored at ⫺70 °C. Lipids were measured at the Collaborative Studies Clinical Laboratory (Fairview-University Medical Center, Minneapolis, MN) within 2 weeks of sample collection, using CDC/NHLBI (Centers for Disease Control and Prevention/CDC/NIH Lipid Standardization Laboratory) Lipid Standardization Program standards. HDL cholesterol was measured using the cholesterol oxidase method (Roche Diagnostics) after precip2

Clinical Chemistry 62:7 (2016)

itation of non–HDL cholesterol with magnesium/dextran (CV ⫽ 2.9%). LDL cholesterol was calculated by using the Friedewald equation. GlycA. The NMR measurement of GlycA, its chemical origins, and its analytic and biological variability have been described in detail (9 ). NMR spectra used for GlycA analysis were those acquired for NMR LipoProfile testing at the CLIA-certified LipoScience (now LabCorp) clinical laboratory in Raleigh, NC (8 ). GlycA concentrations obtained from archived NMR LipoProfile spectra of baseline plasma from 6751 participants in MESA were used to assess associations with demographic and laboratory parameters, including measures of inflammation. Reported CVs for intra- and interassay precision were 1.9% and 2.6%, respectively (9 ). Inflammatory markers. hsCRP was measured using the BNII nephelometer (Dade Behring). Intra- and interassay analytic CVs ranged from 2.3% to 4.4% and 2.1% to 5.7%, respectively. IL-6 was measured by ultrasensitive ELISA (Quantikine HS Human IL-6 Immunoassay, R&D Systems). The laboratory analytic CV for this assay was 6.3%. D-dimer was measured using an immunoturbidimetric assay on a Sta-R analyzer (Liatest D-DI, Diagnostica Stago). The lower limit of detection was 0.01 ␮g/mL. Intraassay CV for D-dimer was 12.2% and the interassay CV range was 5%–14%. Outcome events: total death, CVD, ChrIRD, and total cancer. As previously described (16 ), participants were contacted at 9-month intervals to identify deaths and hospitalizations, with further reference to the death certificate and the National Death Index. We adjudicated death and hospital records to classify ChrIRD, as previously described in detail (16 ). We excluded diagnoses of injury, acute organ failure, psychoses, substance abuse, and some metabolic disorders such as diabetes mellitus, similar to previous studies. Conceptually, ChrIRD reflected fatal events or those likely to result in intensive care, which included a chronic inflammatory, oxidative stress, or infectious component as one predominant pathophysiology. Two physicians independently identified the inflammatory-related International Classification of Diseases (ICD) codes among non-CVD and noncancer disorders and classified each code independently of any others in the record, achieving substantial agreement with each other, as previously described (16 ). The following chronic and severe pathologic conditions contributed to ChrIRD: chronic infectious diseases; endocrine, nutritional, and metabolic disease; nervous system disorders; respiratory system diseases; digestive system diseases; skin diseases; musculoskeletal and connective tissue disorders; genitourinary diseases; and blood disorders. We classified codes one at a time; consequently, a person could be placed in CVD, cancer, and ChrIRD. A complete list of

GlycA, Clinical Outcome, and Death

ICD codes and their frequencies of occurrence was previously published (16 ). CVD events were adjudicated by physicians on the basis of medical records. CVD was defined as myocardial infarction, angina, stroke, transient ischemic attack, heart failure, and peripheral arterial disease and CVD death, as previously described. We carried out separate analyses of total CVD, coronary heart disease (CHD), stroke, and heart failure. Cancer was based on ICD codes. STATISTICAL ANALYSIS

Mean and SD or counts and percentages were calculated for description. We computed Pearson correlation between GlycA and the other inflammatory variables, as well as blood lipids. We performed Poisson regression (incidence density of the earliest qualifying event over a median period of 12.1 years) predicting the following clinical outcome variables from GlycA, IL-6, hsCRP, and D-dimer variables: total death, any CVD, ChrIRD, and total cancer (death or hospitalization). As subsidiary outcomes, we examined subtypes of CVD, namely CHD, stroke, and heart failure. The relative risk per SD was calculated for each inflammatory variable by using a minimal model, including age, race, sex, and clinic as covariates, and a full model, which includes minimal-model covariates plus height, heart rate, systolic and diastolic blood pressure, blood pressure–lowering medications, body mass index, former and current smoking, diabetes, total cholesterol, HDL cholesterol, triglycerides, cholesterol-lowering medication, and estimated glomerular filtration rate (17 ) (⬍60 mL/min/1.73 m2). The minimal and full models do not adjust for other inflammatory variables. The minimal model evaluates the association of each marker after removing confounding influence of demographic variables, while the full model evaluates independence of the corresponding association after removing overlap in other risk variables, some of which may be in a causal pathway linking the inflammatory marker to each outcome variable. We examined full models across quintiles of each of the inflammatory biomarkers as a way to display the shapes of the associations with outcomes. Because many inflammatory biomarkers are higher in women than in men, associations were also evaluated in sex-specific strata; interaction for each marker was tested by 2 sample t-test comparing each inflammatory marker coefficient in males to that of females. Where sex interaction was found, analysis by quintiles of the inflammatory predictor was computed. Besides studying the independent predictive ability of GlycA and the other inflammatory markers, we studied added utility of whole regression models. For each event, using Poisson regression, we estimated the event probability under the base model (covariates only) and under the alternative model (including one or more inflammatory markers). We computed the “reclassification

probability” as the alternative probability—the base probability. Improvement in prediction probability (IPP) (16 ) for these whole regression equations is defined in 2 ways: first, the differential event rate for those whose reclassification probability was negative (classified to lower risk status) compared to those whose reclassification probability was positive (classified to higher risk status), and, second, the P value of the regression coefficient for reclassification probability as a continuous variable. In both methods, we also adjusted for base risk. We considered P ⬍ 0.05 to be noteworthy in general screening of findings. Analyses were performed using PC-SAS, version 9.4 (SAS Institute). Results Mean (SD) of GlycA was 380.7 (61.1) ␮mol/L and of the other 3 inflammatory biomarkers were hsCRP 3.6 (5.3) mg/L, IL-6 1.6 (1.2) pg/mL, and D-dimer 0.4 (0.8) ␮g/ mL. Correlations of hsCRP with GlycA and IL-6 were 0.45 and 0.46 and of GlycA with IL-6 was 0.30, while D-dimer related less strongly to the other 3 inflammatory biomarkers (r ⫽ 0.07– 0.14). GlycA correlations with blood lipids were 0.15 with total cholesterol, 0.26 with triglycerides, and ⫺0.07 with HDL cholesterol. Other correlates of GlycA (Table 1) were female sex, other than Chinese race/ethnicity, higher heart rate, higher body mass index, higher systolic blood pressure and antihypertensive therapy, more current smokers, more diabetes, more cholesterol-lowering medications, and higher total cholesterol and triglycerides and estimated glomerular filtration rate ⬍60 mL/min/1.73 m2 (Table 1). Relative risks per SD of the inflammatory predictors GlycA, hsCRP, IL-6, and D-dimer for the several outcome variables total death, any CVD, ChrIRD, and total cancer are all shown in Table 2. Event rates by quintiles of the inflammatory biomarkers are shown in Fig. 1. There were 915 total deaths. GlycA, IL-6, and D-dimer were all significantly predictive for total death in all models, including those in which the inflammatory markers were adjusted for each other. Relative risks in the full model run separately for each biomarker were 1.07– 1.20 per SD for each of the variables. hsCRP was significant for total death only in the models that did not include the other inflammatory biomarkers. In these models, examination of quintiles suggests a linear increase for IL-6 and D-dimer, but possibly a threshold relation for GlycA (Fig. 1). Findings were similar regarding the relative risks for any CVD (n ⫽ 922), except that D-dimer was not predictive after adjustment for the other inflammatory biomarkers. Secondary CVD outcomes are given in Table 3 and Table 4 and Fig. 2. For CHD (n ⫽ 509), only GlycA was predictive in the fully adjusted model, while for stroke (n ⫽ 225) only IL-6 was predictive in the full Clinical Chemistry 62:7 (2016) 3

Table 1. Patients’ characteristics in function of GlycA quintiles.a Q1 No. of patients GlycA mean, μmol/L

Q2

Q3

Q4

1297

1336

1325

1288

302

346

377

411

Q5

P trend

1277 472

(Minimum, maximum)

(205,328)

(329,361)

(362,392)

(393,432)

(432,788)

Age, years

61.6 (10.5)

62.3 ± 10.3

62.4 ± 10.3

62.2 ± 10.1

62.1 ± 10.1

0.0802

Male (%)

62.8

54.6

47.7

42

27.7