Early Glycemic Control andMagnitude of HbA1c ReductionPredict Cardiovascular Eventsand Mortality: Population-BasedCohort Study of 24,752 MetforminInitiatorsDiabetes Care 2017;40:800–807 | https://doi.org/10.2337/dc16-2271

OBJECTIVE

We investigated the association of early achieved HbA1c level and magnitude ofHbA1c reductionwith subsequent risk of cardiovascular events or death in patientswith type 2 diabetes who initiate metformin.

RESEARCH DESIGN AND METHODS

This was a population-based cohort study including all metformin initiators withHbA1c tests in Northern Denmark, 2000–2012. Sixmonths after metformin initiation,we classified patients by HbA1c achieved (<6.5% or higher) and by magnitude ofHbA1c change from the pretreatment baseline. We used Cox regression to examinesubsequent rates of acute myocardial infarction, stroke, or death, controlling forbaseline HbA1c and other confounding factors.

RESULTS

We included 24,752 metformin initiators (median age 62.5 years, 55% males)with a median follow-up of 2.6 years. The risk of a combined outcome eventgradually increased with rising levels of HbA1c achieved compared with a targetHbA1c of <6.5%: adjusted hazard ratio (HR) 1.18 (95% CI 1.07–1.30) for 6.5–6.99%,HR 1.23 (1.09–1.40) for 7.0–7.49%, HR 1.34 (1.14–1.57) for 7.5–7.99%, and HR 1.59(1.37–1.84) for‡8%.Resultswere consistent for individual outcomeevents and robustby age-group and other patient characteristics. A large absolute HbA1c reduction frombaseline also predicted outcome: adjusted HR 0.80 (0.65–0.97) for D = 24, HR 0.98(0.80–1.20) for D =23, HR 0.92 (0.78–1.08) for D =22, and HR 0.99 (0.89–1.10) forD =21 compared with no HbA1c change (D = 0).

CONCLUSIONS

A large initial HbA1c reduction and achievement of low HbA1c levels within6 months after metformin initiation are associated with a lower risk of cardiovas-cular events and death in patients with type 2 diabetes.

1Department of Clinical Epidemiology, Instituteof Clinical Medicine, Aarhus University Hospital,Aarhus, Denmark2Novo Nordisk Scandinavia AB, Ørestad, Copen-hagen, Denmark

Corresponding author: Reimar W. Thomsen,rwt@clin.au.dk.

Received 24 October 2016 and accepted 6March2017.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc16-2271/-/DC1.

This article is featured in a podcast available athttp://www.diabetesjournals.org/content/diabetes-core-update-podcasts.

E.S.B. is currently in general practice in Hov,Municipality Søndre Land, Norway.

© 2017 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More infor-mation isavailableathttp://www.diabetesjournals.org/content/license.

Elisabeth Svensson,1

Lisbeth M. Baggesen,1 Søren P. Johnsen,1

Lars Pedersen,1 Helene Nørrelund,1

Esben S. Buhl,2 Christiane L. Haase,2 and

Reimar W. Thomsen1

800 Diabetes Care Volume 40, June 2017

CARDIOVASC

ULA

RANDMETABOLICRISK

Lowering glycated hemoglobin (HbA1c)levels to ,7% (,53 mmol/mol) in mostadults with type 2 diabetes has been arecommended target in treatment guide-lines for more than a decade (1–4) be-cause of the documented effect inreducing microvascular complications(5). In contrast, it remains debatedwhether even tighter glucose control(such as HbA1c,6.5%)may bemore ben-eficial or harmful (6) and what the trueeffect of tight early glucose control is onsubsequent cardiovascular disease (7). Al-though the large clinical randomized con-trolled trials have failed to show a clearbeneficial effect of early intensive glyce-mic control on cardiovascular events intype 2 diabetes (5), long-term follow-upstudies from the UK Prospective DiabetesStudy (UKPDS) (8) and Veterans study (9)have suggested a beneficial cardiovascu-lar effect, termed “metabolicmemory” or“legacy effect.” (10) In observational re-search, most studies, but not all, suggestthat having a low glycemic level mea-sured at some point of time is associatedwith fewer cardiovascular events andmortality in type 2 diabetes (11–18).Some studies found a linear relationshipbetween successively lower glycemic lev-els and fewer cardiovascular events(11,12,18), whereas others reported aJ- or U-shaped curve (14–16). Comparisonof these studies is hampered by inclusionof case patients with prevalent diabeteswith different time of diabetes durationand by different ways of measuring glyce-mic control. Only two of the observationalstudieson cardiovascular risk began at thediagnosis of diabetes (15,18). Olsson et al.(18) found an overall increased risk ofacute myocardial infarction with a time-updated HbA1c ,6.0% (42 mmol/mol)compared with 6–7%, whereas Ostgrenet al. (15) reported the lowest cardiovas-cular risk was seen with an early achievedHbA1c level of 6.8% (51 mmol/mol).Whether the magnitude of early HbA1creduction predicts subsequent prognosisremains unknown.Because current guidelines empha-

size the importance of tight glycemiccontrol early after diabetes diagnosis,before complications have occurred(4), further research is warranted onthe association of early glycemic controlwith cardiovascular events, taking intoaccount HbA1c at the initiation of metfor-min use. Thus, using real-life data fromDanishmedical registries, we investigated

the association of achieved HbA1c leveland magnitude of HbA1c reduction withsubsequent risk of myocardial infarction,stroke, and death in a population-basedcohort of patients with incident type 2diabetes initiating metformin.

RESEARCH DESIGN AND METHODS

SettingWe conducted this cohort study usingdata from existing population-basedmedical registries covering NorthernDenmark region’s 1.8 million residents(30% of Denmark’s population) during2000–2012. Linkage of all registries wasmade possible through the unique CivilPersonal Registrationnumber (19). InDen-mark, most cases of type 2 diabetes arediagnosed by general practitioners (GPs),and anestimated 80%are also treated andmonitored there. The remaining 20% arereferredby theirGP todiabetes outpatientspecialist clinics at public hospitals (20). Allprescribed medications can only be re-deemed at community pharmacies (21).This study was approved by the DanishData Protection Agency (Record number1-16-02-1-08). Because this registry-basedstudy did not involve patient contact, noseparate permission from the Danish Sci-entific Ethical Committeewas required ac-cording to Danish Legislation.

Study PopulationOverall methods are described in moredetail in previous reports (22,23). Inbrief, we identified all patients withtype 2 diabetes aged 30 or older livingin Northern Denmark who initiated first-time ever glucose-lowering drug treat-ment between 1 January 2000 and31 December 2012. Patients had tohave at least one available HbA1c mea-surement within 12 months before andagain ;6 months after treatment initia-tion (if several HbA1c measurements oc-curred during day 60–180, the valueclosest to day 180 was chosen) (n =38,418) (22). Among these individuals,we selectedall patientswith incidentmet-formin monotherapy use (n = 24,752).Information on prescriptions was ob-tained via the Aarhus University Prescrip-tion Database, which has held completedata coverage in Northern Denmark since1998 (21). Information onHbA1cmeasure-ments was obtained via the clinical labo-ratory information system researchdatabase (LABKA), which has held com-plete data on biochemistry test results

from all GPs and hospitals in the regionsince 2000 (24).

We also identified all metforminmonotherapy initiators during 2000–2012 who did not have HbA1c valuesmeasured bothwithin 12months beforeand ;6 months after metformin initia-tion, and these patients were the non-measurement cohort (n = 17,142).

Early Glycemic Control: AchievedHbA1c and Magnitude of HbA1c

ChangeWe defined two ways to measure earlyglycemic control:

1. HbA1c level (%) achieved 6 monthsafter the index prescription of met-formin: ,6.5, 6.5–6.9, 7.0–7.4, 7.5–7.9, $8.

2. Magnitude of HbA1c reduction, definedas the absolute change in HbA1c (%)from baseline before metformin initia-tion to the level achieved at 6 monthsafter prescription: D 24,23,22,21,0, +1, or +2 and above. In more detail,D24 included reductions of23.5% ormore,D23 included reductions from22.5% to23.4%, D 22 from21.5%to 22.4%, D 21 from 20.5% to21.4%, D 0 from 20.4% to +0.4%,D +1 from +0.5% to +1.4%, and D +2included HbA1c increases of 1.5% ormore.

Cardiovascular Events and MortalityUsing the Danish National Patient Reg-istry (DNPR) (25), we identified all pa-tients hospitalized with myocardialinfarction and stroke. Diagnoses of myo-cardial infarction and stroke have beenpreviously validatedwith predictive valuesof 90% and 79%, respectively (26,27). Weused the Civil Registration System to iden-tify all deaths.We also constructed a com-bined outcome (death and/or myocardialinfarction and/or stroke). All diagnostic co-des used in the study are assembled inSupplementary Table 1. We monitoredall patients from180 days aftermetformininitiation until a cardiovascular event,death, emigration, or 31 December 2012.

CovariatesFrom the medical databases, we ob-tained data on patient characteristicsat 180 days after metformin initiation,potentially associated with both the ex-posure (early glycemic control) and out-come (cardiovascular events and death).We obtained data on 19 major disease

care.diabetesjournals.org Svensson and Associates 801

categories included in the Charlson Co-morbidity Index based on patients’ en-tire hospital contact history within5 years before follow-up start. We sep-arately ascertained contacts for obesity,alcoholism-related disorders, and forany macrovascular or microvascular di-abetes complication, including priorclinical biochemical indication of renaldisease from the DNPR and LABKA (fordefinitions, see Thomsen et al. [23] andSupplementary Table 1).FromthePrescriptionDatabase,weob-

tained information on the redemption ofany other glucose-lowering drug between14 days after the first metformin mono-therapy prescription and until follow-upstart. We also obtained information onany antihypertensive treatment, statins,antiplatelet drugs, and treatment withpsychiatric medications before follow-upstart.From LABKA we obtained information

on the latest baseline HbA1c and on thelatest LDL cholesterol and total cholesterolmeasurement taken within 12 months be-fore follow-up start. We assessed whetherrecommended cholesterol targets forpersons with diabetes were met or not(,2.5 mmol/L for LDL cholesterol and,4.5 mmol/L for total cholesterol) (28).

Statistical AnalysesWe used contingency tables to first de-scribe characteristics for all patients atday 180 after metformin initiation, byachieved HbA1c (%) level (,6.5, 6.5–6.9, 7.0–7.4, 7.5–7.9, $8), and by themagnitude of HbA1c reduction in %(D 24 or more, 23, 22, 21, 0, +1, +2and above). We created a stacked-bargraph figure to illustrate the distributionof achievedHbA1c levels for each baselineHbA1c group (Supplementary Fig. 1).We used Cox regression analyses to

compute crude and adjusted hazard ra-tios (HRs) with 95% CIs to examine theassociation between achieved HbA1c

level and absolute change in HbA1c

from baseline (%) at 6 months, respec-tively, and subsequent risk of cardiovas-cular events and death after 6 months.We adjusted all analyses for the follow-ing covariates (see Table 1 for categori-zation): age, sex, baseline HbA1c beforemetformin start, year of follow-up start,micro- and macrovascular complications,obesity, alcoholism, antiplatelet drugs,statins, antihypertensive drugs, psychiat-ric medications, reached cholesterol

target, and other glucose-lowering therapythanmetformin.We repeated the adjustedanalysis excluding baseline HbA1c.

We repeated all analyses, stratified byage-groups and presence of comorbidityat baseline (healthy vs. comorbidity).Adjusted HRs were also stratified bybaseline HbA1c (%) level (,7.5, 7.5–8.9, and $9) (Figs. 1 and 2). As a sensi-tivity analyses, we repeated our analysiswhile censoring all end points occurringwithin the first 2 years (as potentiallytoo soon for glycemic control to havehad an effect) and another includingonly individuals with at least 5 years offull follow-up.

To adjust for the potential residual orunmeasured effect of socioeconomicstatus (using low education as a proxy),we also did a sensitivity analysis withexternal adjustment (29), where we as-sumed that the relative risk of cardio-vascular events in those with a loweducational level was 1.33 (30) andthat the proportion of people with loweducation was 25% among patients withgood glycemic control (,7%) and 40% inthose with higher HbA1c levels. We usedSAS 9.2 software for all analyses.

RESULTS

Table 1 presents the 24,752 patientswith first-time metformin treatment ac-cording to achieved HbA1c level. Com-pared with those who had an HbA1c

of $8% at 6 months, patients whoachieved HbA1c ,6.5% were to a largerextent older ($70 years: 27% vs. 17%),female (47% vs. 38%), and more likely tohave initiated metformin in the most re-cent study years (started in 2010–2012:60% vs. 35%). They also had slightlymore macrovascular (15% vs. 12%) andmicrovascular (24% vs. 21%) complica-tions at baseline, received more preven-tive medications, and had less medicalobesity (8% vs. 11%). Patients withHbA1c,6.5% had amuch lower baselineHbA1ccomparedwith thosewhoattainedavalue $8%, and more of them had re-ceived no further glucose-lowering add-ontherapy within the first 180 days (Table 1).Supplementary Fig. 1 shows the distribu-tion of HbA1c achieved for each baselineHbA1c group.

Table 1 also reports patient charac-teristics by magnitude of HbA1c change.On the one hand, comparedwith patientswith small HbA1c reductions, those with alarge reduction tended to be younger,

had a lower prevalence of macrovascu-lar complications, had less comorbidity,and were prescribed less preventivemedication. On the other hand, patientswith large reductions in HbA1c had highHbA1c at baseline (e.g., 93% of thosewith HbA1c reduction of D 24% had abaseline HbA1c .10%) and received ad-d-on glucose-lowering therapy to agreater extent. In contrast, increasingHbA1c after metformin start was un-usual and indicated young age, use ofpsychiatric medications, and use ofdrugs other than metformin ratherthan having a specific level of baselineHbA1c (Table 1).

Median follow-up for our cohort was2.6 years (interquartile range [IQR] 1.2–4.7). We observed 439 incident myocar-dial infarctions, 594 strokes, and 1,845deaths. Figure 1 shows the unadjustedcumulative incidence of a combinedoutcome event by achieved early glyce-mic level at 180 days. The incidence wasconsistently lower in patients who hadachieved HbA1c ,6.5% and was highestin patients who had an HbA1c $8% at180 days after metformin. As a point ofcomparison, the outcome incidence inindividuals with no available HbA1c testboth before and after metformin (Fig. 1,black stippled line) was similar to that inpatients who achieved a low HbA1c.

Figure 2 shows that after adjustmentfor confounders, the risk of the compos-ite end point increased with rising levelsof early achieved HbA1c, compared withachievement of HbA1c ,6.5%. The ad-justed HR for the combined outcomewas 1.18 (95% CI 1.07–1.30) for anHbA1c level of 6.5–6.99%, HR 1.23(1.09–1.40) for 7.0–7.49%, HR 1.34(1.14–1.57) for 7.5–7.99%, and HR 1.59(1.37–1.84) for $8% (Fig. 2 and Supple-mentary Table 2). Differences betweencrude and adjusted HRs increased in thehighest HbA1c groups; that is, becausepatients with high HbA1c were youngerand had less comorbidity at baseline,their high cardiovascular risk furtherincreased after adjustment for thesedifferences in prognostic factors. Theadjusted model not including baselineHbA1c showed consistent results formyocardial infarction, stroke, and mor-tality but with less precise risk estimates(Supplementary Table 2). Results werealso consistent within different strata ofage and presence or absence of comor-bidity at baseline (Supplementary Fig. 2).

802 Early Glycemic Control and Type 2 Diabetes Outcomes Diabetes Care Volume 40, June 2017

Table

1—Characte

risticsat6monthsafte

rmetfo

rmin

start

among24,75

2metfo

rmin

initia

tors,

acco

rdingto

ach

ieve

dHbA1cormagnitu

deofabso

lute

HbA1creductio

nfro

mbase

line

Achieved

HbA1clevel6

monthsafter

metfo

rmin

startMagn

itudeofHbA1cchange

from

baselin

eto

6months

,6.5%

6.5–6.99%

7.0–7.49%

7.5–7.99%

$8%

Total

24%

ormore

23%

22%

21%

0%+1%

+2%ormore

n=11,849

(48%)

n=6,578

(27%)

n=3,035

(12%)

n=1,434

(6%)

n=1,856

(8%)

N=24,752

(100%)

n=1,893

(8%)

n=1,880

(8%)

n=2,685

(11%)

n=6,590

(27%)

n=11,179

(45%)

n=429

(2%)

n=96

(0%)

Variab

len

%n

%n

%n

%n

%N

%n

%n

%n

%n

%n

%n

%n

%

Baselin

eHbA1c ,%

,6.5

3,99834

2223

271

71

131

4,26717

00

00

30

4156

3,71538

9823

3630

6.5to

,6.9

3,79232

1,95030

2518

473

241

6,06425

00

00

251

2,15029

3,77639

9523

1815

7to

,7.4

1,0969

1,44522

54818

846

493

3,22213

00

10

873

1,67323

1,38314

6415

1412

7.5–7.9

6966

94014

56919

20114

1347

2,54010

00

91

39315

1,53721

5045

7919

1815

8–8.9

7236

80512

69923

40728

36120

2,99512

100

25616

1,09443

1,24617

3073

5513

2722

9–9.9

4544

4136

35912

25718

38321

1,8668

1857

63640

65926

3104

561

143

65

$10

1,0909

80312

58219

43130

89248

3,79815

2,66193

67443

28911

1162

380

174

33

Age-gro

up,years

#60

4,67740

2,51738

1,29643

73651

1,07758

10,30342

1,05456

1,05456

1,33650

2,52038

4,05436

22152

6467

61–70

3,92833

2,12032

95532

42930

46825

7,90032

55029

53228

76629

2,14533

3,77034

11727

2021

.70

3,24427

1,94130

78426

26919

31117

6,54927

28915

29416

58322

1,92529

3,35530

9121

1213

Male

sex6,325

533,594

551,713

56844

591,144

6213,620

551,239

661,184

631,653

623,651

555,584

50253

5956

58

Calen

dar

year

2000–200

3540

5356

5306

10178

12241

131,621

7228

12264

14296

11445

7338

344

106

6

2004–200

61,131

10762

12506

17276

19319

172,994

12500

18296

19429

17975

13712

764

1518

15

2007–200

93,127

262,170

331,147

1543

38643

357,630

31333

18334

18481

18888

14882

859

1417

18

2010–201

27,051

603,290

501,076

36437

31653

3512,507

51615

33618

33889

332,275

353,048

27151

3534

35

Macro

vascular

complicatio

ns

1,77415

1,05916

50217

21215

22912

3,77615

1548

21312

38214

1,07016

1,86617

7618

1516

Micro

vascular

complicatio

ns

2,84824

1,66125

78126

32623

38621

6,00224

33618

37320

59722

1,75427

2,79925

12028

2324

Hospital-d

iagnosed

obesity

8858

4747

2328

1299

20011

1,9208

1377

1337

2299

4998

8538

4911

2021

CharlsonCom

orbidityIndex

level$1

2,49621

1,48623

68823

31822

35919

5,34722

27314

34618

58522

1,49623

2,51623

10224

2930

Antip

lateletdrugs

4,57539

2,76442

1,23741

53337

58432

9,69339

54029

55830

94635

2,78142

4,66342

17140

3435

Statins

7,97067

4,53969

1,96665

86260

1,02555

16,36266

95751

1,00453

1,56258

4,42367

8,09172

27464

5153

Antih

yperten

sivedrugs

8,76374

4,99176

2,23874

98269

1,20765

18,18174

1,12460

1,19063

1,82368

5,05777

8,61277

31173

6467

Psych

iatricmed

ications

2,32920

1,14517

52717

22716

32718

4,55518

29316

29916

47018

1,14317

2,21020

10424

3638

Single

metfo

rmin

at6months

11,05793

6,07692

2,72490

1,21585

1,53683

22,60891

1,40874

1,49780

2,32086

6,10993

10,80897

39091

7679

care.diabetesjournals.org Svensson and Associates 803

The clearest association between higherHbA1c and worse outcomes was ob-served in patients aged 70 years or older.Figure 3 shows that the magnitude of

HbA1c change predicted outcome as well.Large HbA1c reductions were associatedwith the greatest outcome risk reductions

among patients with a high baseline HbA1c(i.e., .9%), although statistical precisionwas limited. In patientswith a lowbaselineHbA1c (i.e.,,7.5%), the pattern tended tobe U-shaped, with HbA1c reductions corre-sponding to D = 22% (from 21.5% to22.4%) predicting increasedoutcome risk.

The overall adjusted HR for a com-bined outcome was 0.80 (95% CI 0.65–0.97) forD =24, HR 0.98 (0.80–1.20) forD =23, HR 0.92 (0.78–1.08) for D =22,and HR 0.99 (0.89–1.10) forD =21 com-pared with the reference group withno HbA1c change (D = 0) (Fig. 3 andSupplementary Table 3). An increasedoutcome risk was seen in patients withincreasing HbA1c despite metformininitiation.

A sensitivity analysis that censored allend points occurring within 2 years andwas restricted to patients with at least5 years of full follow-up (SupplementaryTables 4 and 5 and Supplementary Figs.3 and 4) showed consistent results forbothHbA1c reductionandHbA1c achieved,albeit statistical precision was poorer.

When we externally adjusted for un-measured confounding resulting fromsocioeconomic status (proxy education)for the association between glycemiccontrol and macrovascular events, theanalyses yielded results consistent withthe overall findings.

CONCLUSIONS

In this population-based study of 24,752metformin initiators, attaining a stringentHbA1c goal of,6.5%within 6monthswas

Figure 1—Combined outcome event (acute myocardial infarction, stroke, or death) by achieved early glycemic level. The cumulative incidences of acombined outcome event by achieved HbA1c level 6 months after metformin start are shown among 24,752 metformin initiators in NorthernDenmark during 2000–2012. As a point of comparison, the black stippled line shows the event rate among 17,134 metformin initiators in whom anHbA1c measurement was missing before and after metformin initiation. Time (years) to cardiovascular disease (CVD) or death was from 6 months(180 days) after metformin initiation.

Figure 2—Adjusted HRs for combined outcome event (acute myocardial infarction, stroke, ordeath) by achieved HbA1c 6 months after metformin initiation. HRs are stratified by baselinepretreatment HbA1c levels of ,7.5%, from 7.5 to ,9.0%, and $9.0%. CVD, cardiovasculardisease; REF, reference (HR associated with target ,6.5%).

804 Early Glycemic Control and Type 2 Diabetes Outcomes Diabetes Care Volume 40, June 2017

associated with lower risk of cardiovascu-lar events and mortality, with the riskgradually increasing at higher HbA1c lev-els. A large magnitude of HbA1c reductionsimilarly was associated with a lower sub-sequent risk of adverse outcomes.Our results corroborate findings from a

few previous observational studies onearly glycemic control showing loweredrisk of macrovascular events (15,18),and also in the UKPDS trial (31), all inpatients with newly diagnosed type 2 di-abetes. Olsson et al. (18), monitoring101,799 patients from the Clinical Prac-tice Research Data Link in the U.K. be-tween 1995 and 2011 from start oftype 2 diabetes diagnoses, found an in-creased risk for myocardial infarction of;60% (HR 1.6) at an HbA1c of 7–8% vs.6–7%, with a median follow-up of 5.4years (18). Our corresponding findingwas slightly lower (HR 1.5 [95% CI 1.0–2.1]) comparing HbA1c of$8% vs.,6.5%,with a median follow-up of 2.6 years. Thesimilar findings, despite different catego-rization of HbA1c and use of differentHbA1c measure (updated means vs. earlyglycemic control), strengthens the validityof our results.Our data show that achievement of

stringent glycemic levels is possible inreal life in at least some elderly patientswith comorbidities and that reachingsuch levels predicts lower risk of vascularevents and death. Of note, elderly pa-tients with complications who attained

stringent HbA1c levels in our observa-tional study may have been selected bycaregivers through criteria that are notwell described in our data, for example,by having a low risk of hypoglycemia orhigh patient motivation and self-care (3).In accordance, the updated statementfrom the American Heart Associationand the American Diabetes Associationsuggests stringent targets for selected in-dividual patients, including patientswith a short disease duration (7).

A novel finding in our study was thatthe magnitude of early HbA1c reductionis an independent predictor of lowercardiovascular risk and death, also aftertaking pretreatment HbA1c level into ac-count. Nonetheless, in the subgroup ofpatients who had a low baseline HbA1c(,7.5%) before metformin initiation, areduction in the order of22% tended tobe associated with worse outcomes,consistent with findings from random-ized trials (5). As a result of the statisticalvariation and imprecision of outcomeHRs associated with limited size of thechange in HbA1c subgroups, our findingsshould be interpreted with caution.

Overall, our findings suggest thatrapid glycemic response may be usedas a possible source of identification of asubgroup of patients with a lower risk ofadverse outcomes. It is possible that rapidglycemic responders to therapy initiation(i.e., more easy-to-treat patients) mayhave a different pathological trajectory

and a milder variant of type 2 diabetesthan patients who are poor responders.It is also possible that it is the youngpatients with type 2 diabetes withoutcomplications that clinicians dare totreat intensively; for example , reduceHbA1c.10% to,7%, supported by a re-cent study fromDenmark (32). However,we observed a clear association betweenearly control and improved outcomes alsowhen restricted to people older than70 years in our study. Moreover, we havepreviously observed that young patients,to a lesser extent than older patients, reachan early glycemic control ,7% within6 months, possibly related to the fact thatthe pretreatment HbA1c with young debutof diabetes often is high (33).

The strengths of this study include apopulation-based design within thecomprehensive Danish public healthcare system, and accordingly, our datareflect actual clinical practice in diabetescare. Carstensen et al. (34) showed ahigh sensitivity and positive predictivevalue (.95%) for identifying patientswith type 2 diabetes using Danish regis-tries, with GP registration as the goldstandard. Positive predictive values forimportant comorbidities are also docu-mented as being high in the DNPR (25).Furthermore, we have a comprehensiveassessment of cardiovascular eventsand death, and validity of these codesare also high (26,27). We only looked atmyocardial infarction, stroke, and death,yet other studies have found similar associ-ations with glycemia for heart failure, assummed in the meta-analysis by Erqouet al. (35) Finally, we only included metfor-min initiators, increasing the homogeneityof the population studied and representingthe clinical practice today with metforminas the preferred initial pharmacologicalagent for type2diabetes (4).Whether earlyglycemic control by other oral glucose-lowering drugs is associated with similarbenefits remains to be proven.

Study limitations included that onlyapproximately half of all potentially eli-gible patients had HbA1c measurementswithin the right time frame around ther-apy start. The fact that patients with noHbA1c measurement available had lowoutcome risks suggests that this sub-groupmight have been less severe at start.Over time, the frequency of HbA1c mea-surements has increased, and becausetreatment guidelines have changed overtime from initial lifestyle modification to

Figure 3—Adjusted HRs for combined outcome event (acute myocardial infarction, stroke, ordeath) by magnitude of HbA1c reduction from baseline to 6 months. HRs are stratified bybaseline pretreatment HbA1c levels of ,7.5%, from 7.5 to,9.0%, and $9.0%. CVD, cardiovas-cular disease; REF, reference (HR associated with no change in HbA1c).

care.diabetesjournals.org Svensson and Associates 805

emphasizing early drug treatment, it islikely that patients had high HbA1c levelsin the beginning of our study period pri-marily as a result of late initiation ofmedical therapy and that HbA1c mea-surements from recent years more re-liably reflect baseline glycemic controlin newly diagnosed type 2 diabetes.Nonetheless, analyses stratified by pe-riod of metformin initiation showedconsistent results. Moreover, prescrip-tion redemption is only a marker of ac-tual drug consumption. We also have tobear in mind that the guidelines havechanged over time to encompass individ-ualized therapy (3,4). Pretreatment HbA1clevels have decreased substantially overtime in Denmark and in other countries,and achievement of early glycemic controlhas improved (22).By regression analyses and stratification,

we were able to evaluate the effect of arange of possible confounders of the asso-ciation between early glycemic control andcardiovascular events, and interestingly,few differences were seen comparingcrude and adjusted results and in stratifiedanalyses. Also, the analysis adjusted for po-tential unmeasured confoundingbyeduca-tional level showed consistent results. Wehad no data on tobacco smoking but wereable to adjust for a number of smoking-related diseases. Still, imperfectly mea-sured, unmeasured (e.g., BMI, diet, physicalactivity, social support, motivation, andself-care), or unknown factors may haveaffected our risk estimates.In conclusion, these real-world data

provide evidence that not only achieve-ment of early glycemic control but alsothemagnitudeofHbA1c reductionpredictsdecreased risk of cardiovascular outcomesand mortality in metformin initiators, in-dependent of baseline HbA1c levels attreatment initiation. Whereas causality isdifficult to prove in our observationalstudy design, these results provide anearly prediction tool for identification ofpatient subgroups with type 2 diabetesthat have increased risk for cardiovascularcomplications and death.

Acknowledgments. The authors thank biostat-isticians Johnny A. Kahlert and Nickolaj RisboKristensen, both from the Department of Clin-ical Epidemiology, Aarhus University Hospital,for their kind help and advice on data analysesand preparation of figures.Funding.TheDepartmentofClinicalEpidemiologyis a member of the Danish Centre for Strategic

Research in Type 2 Diabetes (DD2), supported bythe Danish Agency for Science (grant no.09-067009 and 09-075724), the Danish Healthand Medicines Authority, and the Danish Di-abetes Association. Project partners are listedon the website at www.DD2.nu.Duality of Interest. E.S.B. has been an em-ployee of Novo Nordisk Scandinavia AB but hastaken a position as a general practice physicianin Søndre Land Kommune, Norway as of 1 Febru-ary 2016. C.L.H. is an employee of Novo NordiskScandinavia AB. This study was partly supportedby a research grant from Novo Nordisk A/S toAarhus University. The Department of Clinical Ep-idemiology is involved in studies with fundingfrom various companies as research grants to(and administered by) Aarhus University, includ-ing the current study. No other potential conflictsof interest relevant to this article were reported.The Department of Clinical Epidemiology at

Aarhus University had control of the data andretained final authority over design, content, andinterpretation of the analyses, as well as the de-cision to submit the manuscript for publication.Author Contributions. E.S., L.M.B., S.P.J.,E.S.B., C.L.H., and R.W.T. contributed to thestudy concept and design. E.S., L.M.B., S.P.J.,L.P., H.N., E.S.B., C.L.H., and R.W.T. analyzed andinterpreted data. E.S., S.P.J., and R.W.T. draftedthe manuscript. L.M.B., S.P.J., L.P., and R.W.T.contributed to data acquisition. L.M.B. and L.P.contributed to statistical analysis. L.M.B., L.P.,H.N., E.S.B., and C.L.H. critically revised themanuscript for important intellectual content.S.P.J., L.P., E.S.B., C.L.H., and R.W.T. supervisedthe study. S.P.J., E.S.B., C.L.H., and R.W.T. ob-tained funding. S.P.J. andR.W.T. provided admin-istrative, technical, or material support. R.W.T. isthe guarantor of this work and, as such, had fullaccess to all the data in the study and takesresponsibility for the integrity of the data and theaccuracy of the data analysis.Prior Presentation. Part of the results werepresented as poster number 1092-P (GeneralPoster Session 11 June and selected for showcaseinaModeratedPosterDiscussione-PosterTheater12 June) at the 76th Scientific Sessions of theAmerican Diabetes Association, New Orleans, LA,10–14 June 2016. Part of the results were pre-sented as oral presentation number 261 at the32nd International Conference on Pharmacoepi-demiology & Therapeutic Risk Management ofthe International Society for Pharmacoepidemiol-ogy, Dublin, Ireland, 25–28 August 2016.

References1. American Diabetes Association. Standards ofmedical care in diabetes. Diabetes Care 2004;27(Suppl. 1):S15–S352. International Diabetes Federation GuidelineDevelopment Group. Global guideline for type 2diabetes. Diabetes Res Clin Pract 2014;104:1–523. Inzucchi SE, Bergenstal RM, Buse JB, et al.Management of hyperglycaemia in type 2 diabe-tes, 2015: a patient-centred approach. Updateto a position statement of the American DiabetesAssociation and the European Association for theStudy of Diabetes. Diabetologia 2015;58:429–4424. American Diabetes Association. Summaryof revisions. In Standards of Medical Carein Diabetesd2016 . Diabetes Care 2016;39(Suppl. 1):S4–S5

5. Hemmingsen B, Lund SS, Gluud C, et al. Tar-geting intensive glycaemic control versus tar-geting conventional glycaemic control fortype 2 diabetes mellitus. Cochrane DatabaseSyst Rev 2013;11:CD0081436. Garber AJ, Newcomer JW, Hennekens CH.Lower A1c targets in type 2 diabetes and in-creased mortality: causal or casual? Am J Med2013;126:1033–10347. Fox CS, Golden SH, Anderson C, et al.; Ameri-can Heart Association Diabetes Committee of theCouncil on Lifestyle and Cardiometabolic Health,Council on Clinical Cardiology, Council on Cardio-vascular and Stroke Nursing, Council on Cardio-vascular Surgery and Anesthesia, Council onQuality of Care and Outcomes Research, Ameri-can Diabetes Association. Update on preventionof cardiovascular disease in adults with type 2 di-abetes mellitus in light of recent evidence: a sci-entific statement from the American HeartAssociation and the American Diabetes Associa-tion. Diabetes Care 2015;38:1777–18038. Holman RR, Paul SK, Bethel MA, MatthewsDR, Neil HA. 10-year follow-up of intensive glu-cose control in type 2 diabetes. N Engl J Med2008;359:1577–15899. Hayward RA, Reaven PD, Wiitala WL, et al.;VADT Investigators. Follow-up of glycemic con-trol and cardiovascular outcomes in type 2 di-abetes. N Engl J Med 2015;372:2197–220610. CefaluWT, Rosenstock J, LeRoith D, Blonde L,RiddleMC.Getting to the “heart”of thematter ondiabetic cardiovascular disease: “thanks for thememory.” Diabetes Care 2016;39:664–66711. Stratton IM, Adler AI, Neil HA, et al. Associ-ation of glycaemia with macrovascular and mi-crovascular complications of type 2 diabetes(UKPDS 35): prospective observational study.BMJ 2000;321:405–41212. Eeg-Olofsson K, Cederholm J, Nilsson PM,et al. New aspects of HbA1c as a risk factor forcardiovascular diseases in type 2 diabetes: anobservational study from the Swedish NationalDiabetes Register (NDR). J Intern Med 2010;268:471–48213. Skriver MV, Støvring H, Kristensen JK,Charles M, Sandbæk A. Short-term impact ofHbA1c on morbidity and all-cause mortality inpeople with type 2 diabetes: a Danish popula-tion-based observational study. Diabetologia2012;55:2361–237014. Nichols GA, Joshua-Gotlib S, ParasuramanS. Glycemic control and risk of cardiovasculardisease hospitalization and all-cause mortality.J Am Coll Cardiol 2013;62:121–12715. Ostgren CJ, Sundstrom J, Svennblad B,Lohm L, Nilsson PM, Johansson G. Associationsof HbA1c and educational level with risk of car-diovascular events in 32,871 drug-treated pa-tients with type 2 diabetes: a cohort study inprimary care. Diabet Med 2013;30:e170–e17716. Chiang HH, Tseng FY, Wang CY, et al. All-cause mortality in patients with type 2 diabetesin association with achieved hemoglobin A(1c),systolic blood pressure, and low-density lipo-protein cholesterol levels. PLoS One 2014;9:e10950117. Kranenburg G, van der Graaf Y, van derLeeuw J, et al.; SMART Study Group. The relationbetween HbA1c and cardiovascular events in pa-tients with type 2 diabetes with and without vas-cular disease. Diabetes Care 2015;38:1930–1936

806 Early Glycemic Control and Type 2 Diabetes Outcomes Diabetes Care Volume 40, June 2017

18. Olsson M, Schnecke V, Cabrera C, Skrtic S,Lind M. Contemporary risk estimates of threeHbA1c variables for myocardial infarction in101,799 patients following diagnosis of type 2diabetes. Diabetes Care 2015;38:1481–148619. Schmidt M, Pedersen L, Sørensen HT.The Danish Civil Registration System as a toolin epidemiology. Eur J Epidemiol 2014;29:541–54920. Thomsen RW, Friborg S, Nielsen JS, SchrollH, Johnsen SP. The Danish Centre for StrategicResearch in Type 2 Diabetes (DD2): organizationof diabetes care in Denmark and supplementarydata sources for data collection among DD2 studyparticipants. Clin Epidemiol 2012;4(Suppl. 1):15–1921. Ehrenstein V, Antonsen S, Pedersen L. Ex-isting data sources for clinical epidemiology:Aarhus University Prescription Database. ClinEpidemiol 2010;2:273–27922. Thomsen RW, Baggesen LM, Svensson E,et al. Early glycaemic control among patientswith type 2 diabetes and initial glucose-loweringtreatment: a 13-year population-based cohortstudy. Diabetes Obes Metab 2015;17:771–78023. Thomsen RW, Baggesen LM, Søgaard M,et al. Early glycaemic control in metforminusers receiving their first add-on therapy: a pop-ulation-based study of 4,734 people with type 2diabetes. Diabetologia 2015;58:2247–2253

24. Grann AF, Erichsen R, Nielsen AG, Frøslev T,Thomsen RW. Existing data sources for clinicalepidemiology: the clinical laboratory informa-tion system (LABKA) research database at AarhusUniversity, Denmark. Clin Epidemiol 2011;3:133–13825. Schmidt M, Schmidt SA, Sandegaard JL,Ehrenstein V, Pedersen L, Sørensen HT. TheDanish National Patient Registry: a review ofcontent, data quality, and research potential.Clin Epidemiol 2015;7:449–49026. Madsen M, Davidsen M, Rasmussen S,Abildstrom SZ, Osler M. The validity of the di-agnosis of acutemyocardial infarction in routinestatistics: a comparison of mortality and hospi-tal discharge data with the Danish MONICA reg-istry. J Clin Epidemiol 2003;56:124–13027. Wildenschild C, Mehnert F, Thomsen RW,et al. Registration of acute stroke: validity in theDanish Stroke Registry and the Danish NationalRegistry of Patients. Clin Epidemiol 2013;6:27–3628. Danish Society of Cardiology. Dyslipidemia.Available fromhttp://nbv.cardio.dk/dyslipidaemi.Accessed 4 October 2016.29. Lash TL, Fox MP, Fink AK. Sensitivity analy-ses for unmeasured confounding. In ApplyingQuantitative Bias Analysis to EpidemiologicalData. Lash TL, Fow MP, Fink AK, Eds. Oxford,Springer-Verlag, 2009

30. Rasmussen JN, Rasmussen S, Gislason GH,

et al. Mortality after acutemyocardial infarction

according to income and education. J Epidemiol

Community Health 2006;60:351–35631. UK Prospective Diabetes Study (UKPDS)

Group. Effect of intensive blood-glucose control

with metformin on complications in overweight

patients with type 2 diabetes (UKPDS 34). Lan-

cet 1998;352:854–86532. Mor A, Berencsi K, Svensson E, et al. Pre-

scribing practices and clinical predictors of glu-

cose-lowering therapy within the first year in

people with newly diagnosed type 2 diabetes.

Diabet Med 2015;32:1546–155433. Svensson E, Baggesen LM, Thomsen RW,

et al. Patient-level predictors of achieving early

glycaemic control in Type 2 diabetes mellitus: a

population-based study. Diabet Med 2016;33:

1516–152334. Carstensen B, Kristensen JK, Marcussen

MM, Borch-Johnsen K. The National Diabetes

Register. Scand J Public Health 2011;39(Suppl.):

58–6135. Erqou S, Lee CT, Suffoletto M, et al. Associ-

ation between glycated haemoglobin and the

risk of congestive heart failure in diabetes mel-

litus: systematic review and meta-analysis. Eur J

Heart Fail 2013;15:185–193

care.diabetesjournals.org Svensson and Associates 807