Do Cardiac BiomarkersNT-proBNP and hsTnT PredictMicrovascular Events in PatientsWith Type 2 Diabetes? ResultsFrom the ADVANCE TrialDOI: 10.2337/dc13-2625

OBJECTIVE

We investigated microvascular event risk in people with type 2 diabetes andassessed whether N-terminal pro–B-type natriuretic peptide (NT-proBNP), andhigh-sensitivity troponin T (hsTnT) improved prediction.

RESEARCH DESIGN AND METHODS

We performed a case-cohort study, including 439 incident cases of microvascularevents (new or worsening nephropathy or retinopathy) and 2,946 noncase sub-jects identified from participants in the Action in Diabetes and Vascular Disease:Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) trial.NT-proBNP and hsTnT were measured in stored plasma samples using automatedcommercial assays.

RESULTS

After adjustment for age, sex, and randomized treatment, the hazard ratios formicrovascular events per 1-SD increase in the log-transformed hsTnT and NT-proBNP were 1.67 (95% CI 1.51–1.85) and 1.63 (1.44–1.84), respectively. Afterfurther adjustment for classical and diabetes-related cardiovascular disease riskfactors, the hazard ratios attenuated to 1.40 (1.24–1.58) and 1.41 (1.24–1.60),respectively. While the C statistic did not improve on addition of hsTnT or NT-proBNP for the total microvascular end point, a combination of both markersimproved the prediction of nephropathy (P = 0.033) but not retinopathy (P =0.72). The corresponding net reclassification indices in a three–risk categorymodel (<10%, 10–15%, and >15% 5-year risk) for all microvascular events were7.31% (95% CI 2.24–12.79) for hsTNT addition, 6.23% (1.74–11.5) for NT-proBNPaddition, and 7.1% (1.5–12.9) for both markers together.

CONCLUSIONS

These data suggest that cardiac biomarkers moderately improve microvascularevent risk prediction, in particular the risk of nephropathy. Further studies exam-ining the value of this approach for trial design and clinical use are warranted.

1Institute of Cardiovascular and Medical Scien-ces, University of Glasgow, Glasgow, U.K.2George Institute for Global Health, University ofSydney, Sydney, Australia3Service d’Endocrinologie Diabetologie Nutri-tion, Groupe Hospitalier Bichat–Claude Bernard,Paris, France4Institute of Cardiovascular Sciences, UniversityCollege London, and the National Institute forHealth Research University College London Hos-pitals Biomedical Research Centre, London, U.K.5Department of Clinical Pharmacology, ImperialCollege London, London, U.K.6School of Mathematical Sciences, University ofTechnology, Sydney, Sydney, Australia7University of Melbourne and Royal MelbourneHospital, Melbourne, Australia

Corresponding author: Naveed Sattar, naveed.sattar@glasgow.ac.uk.

Received 11 November 2013 and accepted 13April 2014.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc13-2625/-/DC1.

© 2014 by the American Diabetes Association.See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.

Paul Welsh,1 Mark Woodward,2

Graham S. Hillis,2 Qiang Li,2

Michel Marre,3 Bryan Williams,4

Neil Poulter,5 Louise Ryan,6

Stephen Harrap,7 Ansuhka Patel,2

John Chalmers,2 and Naveed Sattar 1

Diabetes Care 1

EPIDEM

IOLO

GY/H

EALTH

SERVICES

RESEA

RCH

Diabetes Care Publish Ahead of Print, published online May 30, 2014

Measurement of circulating concentra-tions of B-type natriuretic peptide orN-terminal pro–B-type natriuretic pep-tide (NT-proBNP) and high-sensitivitytroponin assays improve the predictionof cardiovascular disease (CVD) in generalpopulations (1–3). Recent data from theAction in Diabetes and Vascular Disease:Preterax and Diamicron Modified Re-lease Controlled Evaluation (ADVANCE)trial of intensive glucose control andblood pressure lowering also show thatNT-proBNP and high-sensitivity troponinT (hsTnT) improve CVD risk prediction inthose with type 2 diabetes (4). Microvas-cular diseases also cause significantmorbidity in patients with diabetes; forinstance, diabetic nephropathy is a lead-ing cause of end-stage renal disease (5).Clinical risk scores to predict the micro-vascular complications of diabetes arenot in routine use but may be valuableto help clinicians determine which pa-tients are at highest risk and to aid trialdesign.Clearly, there is some interrelationship

between increased microvascular andmacrovascular risk (6).Therefore, puta-tive microvascular risk scores might beexpected to contain some macrovascularrisk factors, as has been recently demon-strated (7). However, the mechanismsunderlying elevation in risk of bothmicro-vascular and macrovascular risk are notclear. Microvascular disease in the myo-cardium might be a major cause of car-diac morbidity among people with type 2diabetes (8,9), and such subtle changes incardiac function (e.g., arising from sub-clinical ischemia) might often precedefrank clinically detectable peripheral mi-crovascular disease. Indeed, elevated na-triuretic peptides in acute coronarysyndromesmaybe correlatedwithabnor-mal perfusion in patients undergoing an-gioplasty (10). Further, cross-sectionaldata in people with diabetes suggest apositive association among cardiac bio-markers, cardiac function, and “periph-eral” microvascular disease (11). Finally,prospective data from ADVANCE haverecently demonstrated that an ele-vated resting heart rate (potentially anearly marker of cardiac overload) isstrongly associated with increased riskof nephropathy and retinopathy (12),and data from Zwolle Outpatient Diabe-tes project Integrating Available Care(ZODIAC)-33 show that copeptin eleva-tion predicts albuminuria (13). Thus, early

changes in cardiac biomarkers may pre-cede presentation with frank microvas-cular disease and yield prognosticinformation that enables intensive inter-ventions to be targeted to those atgreatest risk. Developing a clearer under-standing of any relationship betweencardiac function/metabolism and mi-crovascular end points is importantboth from an aetiological perspectiveand potentially to develop risk scoresto guide clinical management. We thusaimed to 1) investigate the associationbetween biomarkers of cardiac stressand incident microvascular outcomesin patients with type 2 diabetes and 2)investigate the incremental ability ofcardiac biomarkers to predict microvas-cular events.

RESEARCH DESIGN AND METHODS

We performed a nested case-cohortstudy (an efficient prospective study de-sign that has power similar to that of afull cohort study) (14) to investigatethe association between circulating lev-els of NT-proBNP and hsTnT and micro-vascular events in patients with type2 diabetes who participated in theADVANCE study (clinical trial reg. no.NCT00145925, clinicaltrials.gov). Thecase-cohort design has specific advan-tages over a matched nested case-control study, including the ability todevelop risk scores in which the effectof important variables (such as age andsex) are not minimized and the ability toinvestigate several distinct end pointssimultaneously (such as macrovascular[4] and microvascular outcomes).

The ADVANCE study recruited 11,140patients with type 2 diabetes from 215centers in 20 countries between June2001 and March 2003 (12,13). Partici-pants were $55 years of age and hadbeen diagnosed with type 2 diabetes af-ter the age of 30 years. In addition, theywere required to have a history of CVDor one or more additional cardiovascularrisk factors. The studymade two random-ized comparisons: a double-blind assess-ment of the efficacy of perindopril-indapamide (2mg/0.625mg for 3monthsincreasing, if tolerated, to 4mg/1.25mg)versus placebo and an open-label evalu-ation of an intensive glucose-loweringregimen using modified-release glicla-zide, with a target HbA1c of #6.5%(48 mmol/mol) vs. standard, guideline-based glycemic control. The study was

approved by the ethics committee foreach participating center, and all par-ticipants provided written informedconsent. Detailed study methods, in-cluding demographic and clinical mea-surements, and the main results of theADVANCE study have previously beenreported (15,16) (SupplementaryData). Plasma samples were obtainedfrom all study participants at baselineand stored at2808C for a median of 7.8years.

The primary trial outcomes werecomposites of major macrovascularandmicrovascular events (15,16). Majormacrovascular events were cardiovas-cular death, nonfatal myocardial in-farction, or nonfatal stroke. Majormicrovascular events were defined,a priori, as a composite of new orworsening nephropathy or retinopathy.Nephropathy was defined as thedevelopment of macroalbuminuria(i.e., a urinary albumin [micrograms]–to–creatinine [milligrams] ratio (ACR)of .300, confirmed by two results),doubling of the serum creatinine levelto $200 mmol/L (2.3 mg/dL) (with non-qualifying exceptions of acute illnessand subsequent recovery of renal func-tion or terminal illness) (15), the needfor renal replacement therapy due tokidney disease (in the absence of othermedical causes requiring transient dial-ysis), or death due to renal disease. Par-ticipants had their creatinine levelsmeasured as part of the study protocolat baseline, 4 months, and 1 year andannually thereafter until completion ofthe study, with further tests at the dis-cretion of treating clinicians. UrinaryACR was measured as part of the studyprotocol at baseline, 2 years, 4 years,and completion of the study. Glomeru-lar filtration rate was estimated (eGFR)using the Modification of Diet in RenalDisease formula. New or worsening ret-inopathy was defined as the develop-ment of proliferative retinopathy(identified by the incidence of newblood vessels on the disc or elsewhere,vitreous hemorrhage, preretinal hemor-rhage, and fibrous proliferations on thedisc or elsewhere in a participant foundnot to have this condition at entry),macular edema (characterized by a ret-inal thickening within 1 disc diameter ofthe macular center in a participant notfound to have this condition at entry),diabetes-related blindness (corrected

2 Cardiac Biomarkers and Microvascular Disease Diabetes Care

visual acuity 3/60 or worse, persistingfor $3 months and known to not bedue to non–diabetes-related causesin a participant not found to have thiscondition at entry), or the use of retinalphotocoagulation therapy. Participantsunderwent formal eye examinationand visual acuity testing at baseline,2 years, 4 years, and completion of thestudy. Secondary end points were theindividual components of the primaryoutcome (new or worsening nephropa-thy and new or worsening retinopathy).An End Point Adjudication Committee,unaware of treatment allocation, re-viewed source documentation for all in-dividuals who had a suspected primaryend point of the ADVANCE study. In apost hoc analysis, we defined baselinemicrovascular disease as those with pre-vious history of any microvascular event,ACR.300, or eGFR,60 mL/min/1.73m2.Stored plasma sampleswere available

from all countries involved in ADVANCE,except China and India, giving a totalbase population of 7,376. A case-cohortstudy was designed for biomarker mea-surement (Fig. 1).Biomarkers of interest were mea-

sured on stored EDTA plasma samples.NT-proBNP and hsTnT were measuredusing electrochemiluminescence immu-noassays, performed on a Roche Elecsys2010 automated platform (Roche Diag-nostics, BurgessHill, U.K.). TheNT-proBNP

and hsTnT assays have lower detectionlimits of 5 pg/mL and 3 ng/L, respectively(17). For both assays, results below thelimit of detection were reported as 50%of the functional limit of detection foruse in linear models. Assays were per-formed using themanufacturer’s calibra-tors and quality controls. NT-proBNP andhsTnT had assay coefficients of variationof 6.5% and 4.5% for the low control and3.8% and 9.1% for the high control,respectively.

Statistical AnalysisCategorical data are presented as num-ber (percentage) and continuous data asmean (SD), where symmetrically distrib-uted, or median (interquartile range)where skewed. Data for distributionsof risk factors were derived in the over-all subcohort and split by whether theparticipant experienced a primary orsecondary microvascular event duringfollow-up.

Hazard ratios for a 1-SD increase ineach of hsTnT and NT-proBNP, afterlog transformation, on incident micro-vascular events (a statistical modelthat maximizes power and has been pre-viously used in ADVANCE [4] and else-where [1–3]) were obtained fromweighted Cox regression models usingthe STSELPRE procedure for full case-cohort analyses (StataCorp, College Sta-tion, Texas). Nonlinearity was tested by

comparing the deviances of linear andcategorical models and by the inclusionof polynomial components (quadraticand cubic terms). Other analyses wereperformed using SAS, version 9.2 (SASInstitute, Cary, NC). All P values reportedare two sided, with the 5% thresholdused to determine significance.

Three models, with different poten-tial confounding variables, were fittedfor each cardiac biomarker/outcomecombination (Supplementary Data). Inaddition, age, sex, and randomizedtreatment-adjusted survival curveswere modeled by quarters of the distri-bution of the cardiac biomarkers in therandom subcohort only (n = 3,500). Thisrestriction limits the potential biases inprediction potentially caused by taking anonrandom sample.

Prediction metrics for microvascularevents were investigated in the randomsubcohort. C statistics for 5-year risk,accounting for censoring (18), were de-termined for the baseline clinical modeland for this model plus each biomarkerindividually and in combination. TheC statistic has been criticized for insen-sitivities to changes in clinical decisionsyielded for information gained. There-fore, we also used net reclassificationindex (NRI), which estimates correctchanges in clinical classification acrossrisk thresholds (19). We also measuredthe integrated discrimination index and

Figure 1—The ADVANCE trial: the case-cohort study design and data availability.

care.diabetesjournals.org Welsh and Associates 3

the relative integrated discrimination in-dex, which can be considered as a contin-uous version of NRI with probabilitydifferences used instead of categories(19). We estimated the ability of cardiacbiomarkers to appropriately reclassify5-year risk, using these metrics by meth-ods suitable for survival data, with boot-strapping to estimate 95%CI (19). NRIwasderived using a continuous model forchanges in risk classification and from acategorical (or threshold) model basedon 10%, 10–15%, and.15% 5-year risk.

RESULTS

In the subcohort of 3,500 randomlyselected participants, 320 (9.1%)experienced a primary microvascularevent during a median of 5 years’ follow-up. The entire case-cohort study com-prised 3,680 participants (Fig. 1). Ofthese, 3,385 participants had data forplasma NT-proBNP and hsTnT results(92.0%), of whom 439 experienced a mi-crovascular event, and 2,946 were

noncases (Fig. 1). The mean age of thestudy cohort was 67 years, and 61%weremale (Table 1). Median hsTnT was 5 ng/L(64.6% of the cohort had a detectabletroponin), and median NT-proBNP was90 pg/mL. hsTnT concentrations werehigh (.14 ng/L) in 15.4% and NT-proBNPconcentrations high (.400 pg/mL) in12.8% of the case-cohort study. Crudebaseline predictors of incident microvas-cular risk included male sex, increasedduration of diabetes, systolic blood pres-sure, HDL cholesterol, triglycerides, rest-ing heart rate, HbA1c, fasting glucose,ACR, and eGFR (P , 0.01 for all). NT-proBNP and hsTnT were also higheramong case subjects (P , 0.01). Dataare also available for nephropathy andretinopathy end points separately (Sup-plementary Data).

Both NT-proBNP and hsTnT showedbroadly expected associations with riskfactors for CVD and diabetes, as previ-ously reported (4) (SupplementaryData). NT-proBNP and hsTnT levels

were moderately associated with eachother (r = 0.404, P , 0.001)

There was a log-linear association be-tween both biomarkers and primary andsecondary microvascular outcomes(P values ,0.001), with no evidence ofnonlinearity. After adjustment for age,sex, and randomized treatment, bothNT-proBNP and hsTnT associated signif-icantly with the risk for primary micro-vascular events (Figs. 2 and 3; Table 2),with each parameter being morestrongly associated with risk for ne-phropathy than retinopathy. These as-sociations remained but were attenuatedafter adjustment for other clinical risk fac-tors for microvascular events. There wasno strong evidence of an interaction effectfor any of the main adjustment variables(data not shown). The association of quar-ters of both cardiac biomarkers with thenephropathy end point had no interactionwith presence of baseline microvasculardisease (P = 0.155 for hsTnT and P =0.552 for NT-proBNP).

The strongest single predictor of futurenephropathy was ACR; C statistic 0.812(95% CI 0.780–0.845). In contrast, reti-nopathy was most strongly predicted(univariately) by duration of diabetes andHbA1c;C statistics 0.636 (0.595–0.678) and0.604 (0.557–0.650), respectively. A base-line model including all relevant routinelycollected baseline clinical variables (age,sex, randomized treatment, country, du-ration of diabetes, current smoking, sys-tolic blood pressure, BMI, ACR, eGFR,HbAlc, glucose, total and HDL choles-terol, triglycerides, resting heart rate,and history of CVD) discriminated thosewho experienced any microvascularevents moderately well (C statistic =0.731) and the nephropathy end pointvery well (C statistic = 0.829) (Table 3).Addition of hsTnT to the model predict-ing any microvascular complication didnot improve the C statistic (P = 0.201)but improved the continuous NRI (P ,0.001), with the improvement most evi-dent for the prediction of nephropathy.Likewise, NT-proBNP showed some evi-dence of improving the continuous NRI(P , 0.001) but not the C statistic (P =0.153). Improvements in the integrateddiscrimination index were also sig-nificant for both biomarkers (Supple-mentary Data).

Building a categorical NRI for theoret-ical clinical treatment decisions, weused three cutoffs similar to those

Table 1—Baseline characteristics classified by microvascular event outcomestatus

Microvascular events

Overall PYes No

N 439 2,946 3,385

Age (years) 66.5 (6.5) 66.6 (6.6) 66.6 (6.6) 0.86

Male sex 295 (67.2%) 1,729 (58.7%) 2,024 (59.8%) ,0.01

Duration of diabetes(years) 9.8 (6.9) 7.4 (6.2) 7.7 (6.3) ,0.01

Current smokers 61 (13.9%) 444 (15.1%) 505 (14.9%) 0.52

History of CVD 156 (35.5%) 978 (33.2%) 1,134 (33.5%) 0.33

BMI (kg/m2) 30.2 (5.4) 30.1 (5.3) 30.1 (5.3) 0.69

Systolic BP (mmHg) 150.5 (22.0) 146.9 (21.4) 147.3 (21.5) ,0.01

Diastolic BP (mmHg) 81.1 (11.3) 81.9 (10.7) 81.8 (10.8) 0.18

Total chol. (mmol/L) 5.12 (1.10) 5.15 (1.18) 5.15 (1.17) 0.62

HDL chol. (mmol/L) 1.18 (0.34) 1.24 (0.33) 1.23 (0.33) ,0.01

Triglycerides (mmol/L) 1.79 (1.22, 2.56) 1.70 (1.20, 2.32) 1.70 (1.20, 2.36) 0.03

Resting heart rate (bpm) 74.2 (12.7) 72.4 (12.1) 72.6 (12.2) ,0.01

HbAlc (%) 7.79 (1.57) 7.34 (1.37) 7.39 (1.40) ,0.01

HbA1c (mmol/mol) 61.70 (17.15) 56.68 (14.92) 57.33 (15.32) ,0.01

Glucose (mmol/L) 9.1 (3.4) 8.4 (2.6) 8.5 (2.7) ,0.01

ACR (mg/mg) 48.4 (14.1, 132.6) 12.4 (5.9, 30.0) 14.1 (6.2, 38.9) ,0.01

eGFR (mL/min/1.73m2) 67.1 (19.5) 72.6 (16.2) 71.9 (16.8) ,0.01

CRP (mg/L) 1.72 (0.91, 3.52) 1.83 (0.87, 4.06) 1.80 (0.87, 4.03) 0.38

hsTnT (ng/L) 8 (4, 15) 5 (1.5, 10) 5 (1.5, 11) ,0.01

NT-proBNP (pg/mL) 119 (46, 324) 78 (31, 182) 84 (33, 200) ,0.01

log-transformed hsTnT 2.1 (1.4, 2.7) 1.6 (0.4, 2.3) 1.6 (0.4, 2.4) ,0.01

log-transformedNT-proBNP 4.8 (3.8, 5.8) 4.4 (3.4, 5.2) 4.4 (3.5, 5.3) ,0.01

Values are n (%), mean (SD), or median (interquartile range). chol., cholesterol; CRP, C-reactiveprotein.

4 Cardiac Biomarkers and Microvascular Disease Diabetes Care

used for CVD risk prediction (cutoffs of10%, 10–15%, and .15% 5-year risk).Among noncase subjects at baseline,75.2% were at ,10% risk, 12.8% wereat 10–15% risk, and 12.0% were at 15%risk of any microvascular end point. Thecorresponding proportions for case sub-jects were 40.4%, 21.0%, and 38.6%(Supplementary Data). In thismodel, ad-dition of hsTNT, and to a lesser extentNT-proBNP, improved categorical classifi-cation for predictingmicrovascular events(Table 3); categorical NRI was 7.3% (95%CI 2.2–12.8) and 6.2% (1.7–11.5), respec-tively. Neither marker showed anyevidence of improving retinopathy pre-diction in anymodel. There was no strongevidence of incremental benefit fromadding both markers to any model.

CONCLUSIONS

The current study demonstrates that inpatients with type 2 diabetes, higherlevels of NT-proBNP and hsTnT are

associated with an increased risk ofdeveloping microvascular complica-tions, particularly nephropathy. Thisassociation was attenuated after ac-counting for conventional risk factorsfor, and mediators of, macrovascularcomplications; the relationship be-tween NT-proBNP and hsTnT and estab-lished CV risk factors may, at least inpart, explain their prognostic value.Models designed to simulate a clinicalscore for the prediction ofmicrovascularevents showed an ability to discriminatebetween patients who will and will notdevelop microvascular complications.We suggest that subclinical cardiac dis-ease might be an antecedent of periph-eral microvascular disease. Thus, anapproach that combines clinical predic-tors with biomarkers such as NT-proBNPand hsTnT may help to identify patientswith type 2 diabetes who are at greatestrisk ofmicrovascular complications (par-ticularly nephropathy) and may benefit

most from strategies to reduce this risk.However, our data also illustrate that pre-diction of nephropathy and retinopathyare dependent on different risk factors,reflecting distinct mechanistic pathways.As such, specific prediction of retinopathy(without retinal scanning data) remainschallenging using conventional risk fac-tors and cardiac biomarkers.

Potential MechanismsEmerging evidence suggests that func-tional abnormalities in the microvascu-lature can lead to ischemia (20,21),which would be consistent with possiblerises in troponin and NT-proBNP andcompensatory changes in cardiac struc-ture and function. Indeed, several stud-ies have recently investigated anassociation between troponin elevationand microvascular complications aftermyocardial infarction and percutaneouscoronary intervention (22,23). It is im-portant to note that the additional

Figure 2—Kaplan-Meier curve for event-free survival by quarters (Q) of the hsTnT distribution.

care.diabetesjournals.org Welsh and Associates 5

predictive information added by cardiacbiomarkers noted herein occurred de-spite the addition of previous historyof CVD, which is a well-known risk factor(6). Renal function is likely to be an

important determinant of circulatinglevels of biomarkers (24), and as suchthere may be some reverse causality,although this study is prospective andwe have adjusted for baseline eGFR

and ACR. Therefore, although cardiac bio-markers undoubtedly increase with pro-gressive renal dysfunction, that cardiacbiomarkers predict future declinein renal function independently of

Figure 3—Kaplan-Meier curve for event-free survival by quarters (Q) of the NT-proBNP distribution.

Table 2—Hazard ratios (95% CI) for a 1-SD increment in circulating cardiac biomarkers

Microvascular events(N = 3,385, n = 439)

New or worsening nephropathy(N = 3,363, n = 283)

New or worsening retinopathy(N = 3,261, n = 183)

hsTnT (log scale)(1 SD = 1.05)Model 1 1.67 (1.51–1.85); P , 0.001 1.81 (1.60–2.05); P , 0.001 1.48 (1.26–1.73); P , 0.001Model 2 1.40 (1.24–1.58); P , 0.001 1.44 (1.23–1.67); P , 0.001 1.32 (1.09–1.60); P = 0.005Model 3 1.30 (1.15–1.47); P , 0.001 1.31 (1.11–1.54); P = 0.001 1.24 (1.03–1.50); P = 0.026

NT-proBNP (log scale)(1 SD = 1.58)Model 1 1.63 (1.44–1.84); P , 0.001 1.96 (1.67–2.31); P , 0.001 1.30 (1.08–1.56); P = 0.0044Model 2 1.41 (1.24–1.60); P , 0.001 1.52 (1.29–1.80); P , 0.001 1.34 (1.11–1.61); P = 0.002Model 3 1.31 (1.15–1.48); P , 0.001 1.40 (1.17–1.66); P , 0.001 1.26 (1.05–1.51); P = 0.012

N, total number of participants (event and nonevent) in analysis; n, total number of events in analysis. Model 1: adjusted for age, sex, andrandomized treatment allocations. Model 2: additionally adjusted for country, duration of diabetes, current smoking, systolic blood pressure, BMI,ACR, eGFR, HbAlc, glucose, total and HDL cholesterol, triglycerides, C-reactive protein, resting heart rate, and history of CVD. Model 3: additionallyadjusted for the other biomarker (i.e., NT-proBNP/hsTnT).

6 Cardiac Biomarkers and Microvascular Disease Diabetes Care

baseline renal function suggests poten-tial clinical utility. The findings also hintat a potential for cardiac dysfunction tocontribute to renal dysfunction.

Why Predict MicrovascularComplications?Clinical prediction of microvascularevents is not as yet advocated by guide-lines. This is largely due to a lack ofevidence at present regarding evidence-based therapies to prevent onset ofmicrovascular events. ADVANCE re-ported that intensive glucose andblood pressure control reduce the riskof nephropathy (25), and although datafrom the ACCORD (Action to ControlCardiovascular Risk in Diabetes) trialand VADT (Veterans Affairs DiabetesTrial) were less positive, results wereconsistent with reduction or delay inmicroalbuminuria and macroalbuminu-ria (26,27). A recent meta-analysis con-cluded there was inconsistent evidenceto promote intensive interventions toprevent microvascular disease (28). Fu-ture trials in this area would benefit frombetter identification of high-risk patientsto enable more effective evaluation ofinterventions. From our reclassification

model (using the .15% 5-year risk pre-dicted risk cutoff after addition of car-diac biomarkers), 13.2% of noncasesubjects would be in the highest risk cat-egory and 44.9% of case subjects. Ourresults are thus important, since predic-tion of microvascular events, in particu-lar nephropathy, would be useful forfuture trials of new agents designed toprevent or slow progression of nephrop-athy, an area of intense interest and ac-tivity (29). Indeed, there is a raft ofongoing work in the field of proteomicsto improve the prediction of kidney dis-ease (30,31). While -omics based tech-nology is not yet ready for use inroutine biochemistry in terms of stan-dardization, quality control, or technicalpragmatism, our data illustrate that sim-ple clinical biomarkers predict nephrop-athy very well and that other simple andfar less costly blood-based tests alreadyin routine use, such as cardiac bio-markers, offer incremental predictivebenefit. While our data are hypothesisgenerating, hinting at the potential for car-diac biomarkers to enhance prediction ofdiabetic nephropathy, it should be recog-nized that cardiac biomarkers are predic-tive of macrovascular risk (4); therefore,

such biomarkers have the potential to beroutinely assessed for other purposes. Fur-ther studies are required to assess the fullpotential clinical implications of our find-ings for microvascular risk prediction.

Recent data in a case-control studyfrom the Prevention of Renal and Vascu-lar End-stage Disease (PREVEND) hyper-tensive cohort also suggested that hsTnTis associated with incident microalbumi-nuria (32), lending external support toourresults. We expand on these data byshowing associations of NT-proBNP andtroponin with renal and retinopathy endpoints, aswell as building predictionmod-els. The majority of microvascular eventsin ADVANCE, as in most other cohorts ofpeople with diabetes, were renal. It re-mains possible that cardiac biomarkersare capturing information about renalfunction that are not otherwise availableand thus mainly predict renal events.From a risk prediction perspective, how-ever, the additional information gained inpredicting events is nevertheless poten-tially clinically useful.

Strengths and LimitationsThe current prospective study de-scribes a cohort derived from a large

Table 3—Reclassification and discrimination statistics (95% CI) for circulating cardiac biomarkers after inclusion in a modelwith clinical markers of risk*

All microvascular events(N = 3,225, n = 279)

New or worsening nephropathy(N = 3,225, n = 145)

New or worsening retinopathy(N = 3,225, n = 147)

Base C statistic* 0.731 (0.701, 0.762) 0.829 (0.795, 0.864) 0.727 (0.685, 0.769)

hsTnTC statistic† 0.738 (0.707, 0.769);

P = 0.2010.840 (0.808, 0.872);

P = 0.0450.728 (0.685, 0.771);

P = 0.937NRI: continuous 0.3279 (0.2048, 0.4518);

P , 0.0010.3415 (0.1728, 0.5107);

P , 0.0010.2221 (0.0532, 0.3943);

P = 0.010NRI: categorical 0.0731 (0.0224, 0.1279);

P = 0.0060.0549 (20.0156, 0.1259);

P = 0.1300.0407 (20.0075, 0.0911);

P = 0.104

NT-proBNPC statistic† 0.735 (0.704, 0.766);

P = 0.2920.835 (0.801, 0.869);

P = 0.1530.731 (0.689, 0.773);

P = 0.311NRI: continuous 0.3363 (0.2207, 0.4667);

P , 0.0010.3892 (0.2240, 0.5575);

P , 0.0010.1784 (0.0069, 0.3539);

P = 0.044NRI: categorical 0.0623 (0.0174, 0.11485);

P = 0.0020.0561 (20.0142, 0.1292);

P = 0.1240.0663 (0.0113, 0.1228);

P = 0.020

hsTnT and NT-proBNPC statistic† 0.738 (0.707, 0.769);

P = 0.1930.842 (0.809, 0.874);

P = 0.0330.729 (0.686, 0.772);

P = 0.715NRI: continuous 0.3097 (0.1915, 0.4337);

P , 0.0010.4440 (0.2808, 0.6072);

P , 0.0010.1929 (0.0274, 0.3762);

P = 0.028NRI: categorical 0.0712 (0.0154, 0.1293);

P = 0.0140.0795 (0.0058, 0.1588);

P = 0.0300.0183 (20.0331, 0.0724);

P = 0.904

Results for nonmissing data from the random subcohort (N = 3,500). N = total number of participants (event and nonevent) in analysis; n = totalnumber of events in analysis. categories were 10% and 15% 5-year risk. Biomarkers were all analyzed in continuous form after log transformationsfor NT-proBNP and hsTnT. The P value for the C statistic relates to the increase when adding the biomarker to the base model. *Using model 2(without C-reactive protein) described in Table 2. †P value refers to increase in concordance compared with base model.

care.diabetesjournals.org Welsh and Associates 7

population, which was well character-ized and closely followed up. All endpoints were independently adjudicatedaccording to predefined criteria, includ-ing separate reporting of nephropathyand retinopathy end points. This is oneof the first studies to show that subtlesubclinical differences in cardiac bio-markers may run concurrently with oreven precede peripheral microvasculardisease. Prediction of microvascular endpoints is a novel application of the data.Participants in the ADVANCE study, likethose from other randomized controlledtrial populations, represent a selectedcohort, and our results may not be gen-eralizable to all patients with diabetes.For instance, ADVANCE participantswere required to have a history of CVDor CVD risk factors, and although effortsto adjust for these confounding factorswere made, residual confounding re-mains possible. As such, our findingsshould be considered hypothesis gener-ating. Our results relating to clinical riskpredictors are, however, consistent withother data from an ethnically distinctcohort (7), and the baseline character-istics of the ADVANCE cohort are verycomparable with those in several obser-vational studies at the community level(33). Although the large study popula-tion and case-cohort design ensure con-siderable statistical power and allowsreliable correction for many potentialconfounding factors, other possible con-founders may be present, leading to anoverestimate of risk associations. Levelsof hsTnT and NT-proBNP were mea-sured only once, and it would be usefulto address the effects of serial changesin levels in the future. We speculate onthe mechanisms linking cardiac bio-markers to microvascular outcomesbut have no data to examine cardiacperfusion specifically. The NRI modelsthat we have constructed are necessar-ily based on somewhat artificial cutoffs,but they contain most of the pertinentinformation that might typically be usedto estimate the cardiovascular risk ofpatients with diabetes. NT-proBNP andhsTnT were measured on stored serumsamples, and although we cannot ruleout the potential for differential sampledegradation that biases our results,both of these markers appear fairly sta-ble in long-term storage (34,35).In conclusion, we report that clinical

risk scores to predict microvascular

events in patients with type 2 diabetescan be developed using classical CVDand diabetes-related risk factors. Wefurther show that NT-proBNP and hsTnTcan add moderate information to theseclinical risk scores for nephropathyevents. Further work is required to es-tablish whether the prediction of micro-vascular risk can be improved using avariety of biomarkers.

Funding. The ADVANCE trial (clinical trial reg.no. NCT00145925, clinicaltrials.gov) was par-tially funded by the National Health andMedicalResearch Council (NHMRC) of Australia (projectgrant ID 211086 and program grants 358395and 571281). The biomarker work in the currentstudy was funded by the NHMRC of Australia(project grant 632507) and the Diabetes Aus-tralia Research Trust. P.W. is supported byBritish Heart Foundation fellowship FS/12/62/29889. B.W. is a National Institute for HealthResearch (NIHR) Senior Investigator and issupported by the NIHR University CollegeLondon Biomedical Research Centre.Duality of Interest. The ADVANCE trial wasfunded by Servier. M.W. reports receiving con-sulting fees from Roche and lecture fees fromServier and Sanofi. M.M., S.H., and A.P. havereceived lecturing fees from Servier. N.P. reportsreceiving grant support from Pfizer, Julius ClinicalResearch, and Novartis and lecture fees fromGilead, Daiichi-Sankyo, Servier, Takeda, Pfizer,Novo Nordisk, Roche, Boehringer Ingelheim,Medtronic, and Jannsen. J.C. has received re-search grants from Servier as principal investiga-tor for ADVANCE and for the ADVANCE-ONposttrial follow-up study and honoraria fromServier for speaking about these studies atscientific meetings. No other potential conflictsof interest relevant to this article were reported.Author Contributions. P.W. and N.S. per-formed the laboratory analyses and wrote theinitial drafts of the manuscript. M.W. collecteddata, undertook the statistical analyses, and re-vised the drafts of the manuscript for importantscientific content. Q.L. and L.R. undertook thestatistical analyses and revised the drafts of themanuscript for important scientific content. M.M.,B.W., N.P., S.H., A.P., and J.C. collected data andrevised the drafts of the manuscript for importantscientific content. M.W. is the guarantor of thiswork and, as such, had full access to all the data inthe study and takes responsibility for the integrityof the data and the accuracy of the data analysis.Prior Presentation. Parts of this study werepresented in abstract form at the 49th AnnualMeeting of the European Association for theStudy of Diabetes, Barcelona, Spain, 23–27 Sep-tember 2013.

References1. Wannamethee SG, Welsh P, Lowe GD, et al.N-terminal pro-brain natriuretic peptide is amore useful predictor of cardiovascular diseaserisk than C-reactive protein in older men withand without pre-existing cardiovascular dis-ease. J Am Coll Cardiol 2011;58:56–64

2. Welsh P, Doolin O, Willeit P, et al. N-terminalpro-B-type natriuretic peptide and the predic-tion of primary cardiovascular events: resultsfrom 15-year follow-up of WOSCOPS. Eur HeartJ 2013;34:443–4503. Wang TJ, Wollert KC, Larson MG, et al. Prog-nostic utility of novel biomarkers of cardiovas-cular stress: the Framingham Heart Study.Circulation 2012;126:1596–16044. Hillis GS, Welsh P, Chalmers J, et al. The rel-ative and combined ability of high-sensitivitycardiac troponin T and N-terminal pro-B-typenatriuretic peptide to predict cardiovascularevents and death in patients with type 2 diabe-tes. Diabetes Care 2014;37:295–3035. Collins AJ, Foley RN, Herzog C, et al. US RenalData System 2010 Annual Data Report. Am JKidney Dis 2011;57(Suppl. 1):A8, e1–e5266. Elsayed EF, Tighiouart H, Griffith J, et al. Car-diovascular disease and subsequent kidney dis-ease. Arch Intern Med 2007;167:1130–11367. Tanaka S, Tanaka S, Iimuro S, et al.; JapanDiabetes Complications Study Group; JapaneseElderly Diabetes Intervention Trial Group. Pre-dicting macro- and microvascular complicationsin type 2 diabetes: the Japan Diabetes Compli-cations Study/the Japanese Elderly Diabetes In-tervention Trial risk engine. Diabetes Care 2013;36:1193–11998. LaaksoM.Heart in diabetes: amicrovascular dis-ease. Diabetes Care 2011;34(Suppl. 2):S145–S1499. Rosenson RS, Fioretto P, Dodson PM. Doesmicrovascular disease predict macrovascularevents in type 2 diabetes? Atherosclerosis2011;218:13–1810. Rossetti E,MarianiM, Poli A, et al. NT pro-B-type natriuretic peptide levels are related tomicrovascular reperfusion in patients undergo-ing direct percutaneous transluminal coronaryangioplasty for anterior ST-segment elevationmyocardial infarction. J Cardiovasc Med(Hagerstown) 2010;11:359–36411. Grauslund J, Nybo M, Green A, Sjølie AK.N-terminal pro brain natriuretic peptide reflectslong-term complications in type 1 diabetes.Scand J Clin Lab Invest 2010;70:392–39812. Hillis GS, Hata J,WoodwardM, et al. Restingheart rate and the risk of microvascular compli-cations in patients with type 2 diabetesmellitus.J Am Heart Assoc 2012;1:e00283213. Boertien WE, Riphagen IJ, Drion I, et al.Copeptin, a surrogate marker for arginine vaso-pressin, is associated with declining glomerularfiltration in patients with diabetes mellitus(ZODIAC-33). Diabetologia 2013;56:1680–168814. Woodward M. Epidemiology: Study Designand Data Analysis. Boca Raton, Florida, Chap-man & Hall/CRC, 200415. Patel A, MacMahon S, Chalmers J, et al.;ADVANCE Collaborative Group. Effects of a fixedcombination of perindopril and indapamide onmacrovascular and microvascular outcomes inpatients with type 2 diabetes mellitus (theADVANCE trial): a randomised controlled trial.Lancet 2007;370:829–84016. Patel A, MacMahon S, Chalmers J, et al.;ADVANCE Collaborative Group. Intensive bloodglucose control and vascular outcomes in pa-tients with type 2 diabetes. N Engl J Med2008;358:2560–257217. Giannitsis E, Kurz K, Hallermayer K, JarauschJ, Jaffe AS, Katus HA. Analytical validation of a

8 Cardiac Biomarkers and Microvascular Disease Diabetes Care

high-sensitivity cardiac troponin T assay. ClinChem 2010;56:254–26118. Newson RB. Comparing the predictivepower of survival models using Harrells c orSomers’ D. Stata J 2011;10:339–35819. Pencina MJ, D’Agostino RB Sr, SteyerbergEW. Extensions of net reclassification improve-ment calculations tomeasure usefulness of newbiomarkers. Stat Med 2011;30:11–2120. Herrmann J, Kaski JC, Lerman A. Coronarymicrovascular dysfunction in the clinical setting:from mystery to reality. Eur Heart J 2012;33:2771–2782b21. Lanza GA, Crea F. Primary coronary micro-vascular dysfunction: clinical presentation,pathophysiology, and management. Circulation2010;121:2317–232522. Hallen J, Jensen JK, Buser P, Jaffe AS, Atar D.Relation of cardiac troponin I and microvascularobstruction following ST-elevation myocardialinfarction. Acute Card Care 2011;13:48–5123. Mayr A, Klug G, Schocke M, et al. Late mi-crovascular obstruction after acute myocardialinfarction: relation with cardiac and inflamma-tory markers. Int J Cardiol 2012;157:391–39624. Wu AH, Jaffe AS, Apple FS, et al.; NACBWrit-ing Group; NACB Committee. National Academyof Clinical Biochemistry laboratorymedicine prac-tice guidelines: use of cardiac troponin andB-typenatriuretic peptide or N-terminal proB-type

natriuretic peptide for etiologies other than acutecoronary syndromes and heart failure. Clin Chem2007;53:2086–209625. Zoungas S, de Galan BE, Ninomiya T, et al.;ADVANCE Collaborative Group. Combined ef-fects of routine blood pressure lowering andintensive glucose control on macrovascularand microvascular outcomes in patients withtype 2 diabetes: New results from the ADVANCEtrial. Diabetes Care 2009;32:2068–207426. Ismail-Beigi F, Craven TE, O’Connor PJ,et al.; ACCORD Study Group. Combined inten-sive blood pressure and glycemic control doesnot produce an additive benefit on microvascu-lar outcomes in type 2 diabetic patients. KidneyInt 2012;81:586–59427. Duckworth WC, Abraira C, Moritz TE, et al.;Investigators of the VADT. The duration of di-abetes affects the response to intensive glucosecontrol in type 2 subjects: the VA Diabetes Trial.J Diabetes Complications 2011;25:355–36128. Boussageon R, Bejan-Angoulvant T,Saadatian-Elahi M, et al. Effect of intensive glu-cose lowering treatment on all cause mortality,cardiovascular death, and microvascular eventsin type 2 diabetes: meta-analysis of randomisedcontrolled trials. BMJ 2011;343:d416929. Preiss D, Sattar N, McMurray JJ. A system-atic review of event rates in clinical trials in di-abetes mellitus: the importance of quantifying

baseline cardiovascular disease history and pro-teinuria and implications for clinical trial design.Am Heart J 2011;161:210–219, e130. Bonomini M, Sirolli V, Magni F, Urbani A.Proteomics and nephrology. J Nephrol 2012;25:865–87131. Alkhalaf A, Zurbig P, Bakker SJ, et al.;PREDICTIONS Group. Multicentric validationof proteomic biomarkers in urine specific fordiabetic nephropathy. PLoS ONE 2010;5:e1342132. Hellemons ME, Lambers Heerspink HJ,Gansevoort RT, de Zeeuw D, Bakker SJ. High-sensitivity troponin T predicts worsening of al-buminuria in hypertension; results of a nestedcase-control study with confirmation in diabe-tes. J Hypertens 2013;31:805–81233. Chalmers J, Arima H. Importance of bloodpressure lowering in type 2 diabetes: focus onADVANCE. J Cardiovasc Pharmacol 2010;55:340–34734. Agarwal SK, Avery CL, Ballantyne CM, et al.Sources of variability in measurements of car-diac troponin T in a community-based sample:the atherosclerosis risk in communities study.Clin Chem 2011;57:891–89735. Cauliez B, Guignery J, Marinier S, Mariau I,Lavoinne A. Two-year stability of NT-proBNP infrozen samples using the Roche Elecsys system.Ann Clin Biochem 2008;45:318–319

care.diabetesjournals.org Welsh and Associates 9