equal access to health care may diminish the differences in outcome between native and immigrant...

Pediatric Diabetes 2014: 15: 519–527doi: 10.1111/pedi.12157All rights reserved

© 2014 John Wiley & Sons A/S.Published by John Wiley & Sons Ltd.

Pediatric Diabetes

Original Article

Equal access to health care may diminishthe differences in outcome between nativeand immigrant patients with type 1 diabetes

Fredheim S, Delli A, Rida H, Drivvoll A-K, Skrivarhaug T, Bjarnason R,Thorsson A, Lindblad B, Svensson J. Equal access to health care maydiminish the differences in outcome between native and immigrant patientswith type 1 diabetes.Pediatric Diabetes 2014: 15: 519–527.

Background/Objective: Previous studies have found that ethnicity influencesglycemic control. We hypothesized that differences between Nordic andnon-Nordic patients are less pronounced for children with type 1 diabetes inhigh incidence countries in Northern Europe.Research design and methods: We investigated patients aged 0–15 yr innational pediatric registers in Denmark (D), Iceland (I), Norway (N), andSweden (S) (2006–2009). Ethnic origin was defined by maternal country ofbirth as being Nordic or non-Nordic (other countries).Results: The cohort (n = 11,908, 53.0% boys, onset age 7.7 (3.9) yr, diabetesduration 6.1 (3.6) yr, [mean, (SD)]) comprised 921 (7.7%) non-Nordic patients.The frequencies of non-Nordic patients according to country of residencewere: 5.7% (D), 2.7% (I), 5.5% (N), and 9.4% (S). Sex distribution and BMIz-score did not differ between Nordic and non-Nordic patients, butnon-Nordic patients were 0.5 yr younger at onset than Nordic patients(p < 0.0006). Non-Nordic patients had a lower number of daily insulin bolusinjections and higher daily insulin doses compared to their Nordic peers.Patients of non-Nordic origin had slightly higher HbA1c levels(0.6–2.9 mmol/mol, p < 0.001) and, with the exception of Norway, were lessfrequently treated with CSII (p = 0.002) after adjusting for confounders.Conclusions: The reported differences in glycemic regulation between Nordicand non-Nordic type 1 diabetes children and adolescents in four Nordiccountries are diminutive, but persist after accounting for treatment intensity.

Siri Fredheima,b,Ahmed Dellic, Heba Ridaa,Ann-Kristin Drivvolld,Torild Skrivarhaugd,Ragnar Bjarnasone,f,Arni Thorssone,f,Bengt Lindbladg,† andJannet Svenssona,b,†

aDepartment of Pediatrics, HerlevHospital, Herlev,Copenhagen,Denmark; bFaculty ofHealth and Medical Sciences,University of Copenhagen,Copenhagen, Denmark; cDepartmentof Clinical Sciences, LundUniversity/CRC, Skane UniversityHospital SUS, Malmo, Sweden;dDepartment of Pediatric Medicine,Oslo University Hospital, Oslo, Norway;eChildren´s Medical Center, LandspitaliUniversity Hospital, Reykjavik, Iceland;fDepartment of Pediatrics, University ofIceland, Reykjavik, Iceland; andgDepartment of Pediatrics, Institute ofClinical Sciences, The SahlgrenskaAcademy at Gothenburg University,Gothenburg, Sweden†Both authors contributed equally tothe manuscript

Key words: Denmark – ethnicity –Iceland – Nordic – Norway –population register – Sweden –treatment – type 1 diabetes

Corresponding author: SiriFredheim, MD,Department of Pediatrics,Arkaden, Turkisvej 14,Herlev Hospital,DK 2730 Herlev,Denmark.Tel.: +45 3868 5134;fax: +45 3868 5101e-mail: [email protected]

Submitted 5 February 2014.Accepted for publication 1 May 2014

519

Fredheim et al.

Diabetes is one of the most common chronic diseasesin childhood and adolescence with an increasingincidence in Europe and the US (1, 2). Countries inNorthern Europe have the highest incidence of type1 diabetes in the world, with Finland and Swedenpeaking with an incidence of 64 and 44 per 100,000children per year in 2005–2007, respectively (3, 4).The incidence is particularly low in Asia, Africa andSouth America (China and Venezuela, 0.1 per 100,000children per year) (5). The prevalence of type 1 diabetesin a given population varies among other factors byethnicity (6, 7). As the Northern European populationand society are becoming increasingly multicultural,cultural differences are important features of diabetesmanagement. There is evidence that individuals whomigrate from low risk to high risk type 1 diabetes areastend to adopt the risk level for diabetes observed inthe destination area (8), although migration studiesfrom Germany and Sweden report no such risk change(9, 10). Patients with type 1 diabetes from ethnicminorities have been reported to have higher glycatedhemoglobin (HbA1c) than the native population inseveral studies from the US and a higher risk for laterdiabetes complications (11–14).

Explanatory research of differences in ethnicityhas distinguished between genetic, socioeconomic,and sociocultural factors, including health-seekingbehavior and diabetes management. Some arguethat affluence and socioeconomic status (SES) areattributable for the differences in metabolic controlin minority groups (15); whereas, in other studies,ethnicity has been shown to be important (16). Ethnicdifferences in HbA1c were reported in children of immi-grants compared to children of native mothers, despitethe lack of difference in SES in the groups (11, 17).

Recent advances in diabetes care require an evalu-ation in order to optimize treatment in our childhoodpopulation. We examined the influence of ethnic back-ground and treatment regimen on glycemic control ina large Northern European pediatric population.

Methods

Subjects

We conducted a multinational register-based study ofchildren with type 1 diabetes under age 16 yr in theperiod 2006–2009. The patients were drawn from abackground population in Denmark (D), Iceland (I),Norway (N), and Sweden (S) of approximately 20.6million inhabitants (3 April 2013). Exclusion criteriaare described in Fig. 1.

Nordic registers

The study cohort was obtained by combining data fromfour national pediatric diabetes registers in Denmark,

Island, Norway, and Sweden. The nationwide Dan-ish register for childhood diabetes (DanDiabKids) wasestablished in 1996 and has data ascertainment of 99%(18). The register includes records of children diag-nosed with type 1 diabetes from all 18 tertiary pediatricunits across the country. DanDiabKids undergoes acapture–recapture methodology annually. The nation-wide Icelandic registry was established in 1980 and hasextended clinical data from 2002. Data ascertainmentis 100%. All type 1 diabetes incident cases are referredto the clinic in Reykjavik. The nationwide Norwegianregistry for childhood diabetes (Norwegian ChildhoodDiabetes Registry, NCDR) was established in 1989.All pediatric clinics in Norway report hospitalizedincidence cases to this register. Since 2000, NCDRhas additional registered data on diabetes quality carebased on annual systematic examinations with theWorld Health Organization (WHO) Basic InformationSheet. Data ascertainment is 91% (19). The nationwideSwedish register (Swediabkids) is primarily a qualityregister for children with diabetes. Individual data fromall visits to pediatric outpatient clinics in Sweden arecollected from the year 2000. Since 2007, all 43 pediatricdiabetes clinics in Sweden participate. Swediabkidsincludes a special incidence module monitoring all newpatients with diabetes since 2000. By the design ofthe register and in accordance with capture–recapturemethodology, all patients are included.

Data collection

The study period was from 2006 to 2009. To obtaincomparable data between the registers in the respectivecalendar year, the outpatient visit closest to thepatient’s date of birth (D, S) or data from the first visitduring the year (I, N) were extracted for the variablesin question. Of 34759 visits, 30103 were eligible, andthe cohort comprised 11,908 patients. Patients werediagnosed with type 1 diabetes according to WHO cri-teria (20). Baseline (age at visits, sex, date of diagnosis,duration of diabetes, maternal country of birth) andclinical data (weight, length, diabetes treatment, andHbA1c) were extracted from the registers. For Swedishdata, the maternal region of birth was obtained fromStatistics Sweden. Body mass index (BMI) was calcu-lated from weight and height (kg/m2) and standardizedto a standard deviation (SD) score (BMI z-score)according to age and sex, as the participants wereyounger than 18 yr. We used the Centers for DiseaseControl and Prevention (CDC) growth charts for 2000as the reference (21). Data collection was systematicallyhandled at the respective hospital facility. Data werereported to the registers by online registrations (D, I,S) or by completing dedicated paper forms (N) duringfollow-up.

520 Pediatric Diabetes 2014: 15: 519–527

Ethnicity and glycemic control

Fig. 1. Study cohort from four Nordic pediatric diabetes registers and eligible visits and patients. Study cohort, the total number of visitsextracted from all out-patient visits in the registers during 2006–2009. Exclusion criteria and numbers are listed together with the final numberof visits and patients. Danish (DanDiabKids), Icelandic (ICDR), Norwegian (NCDR), and Swedish (Swediabkids) registers.

Ethnic background

We defined ‘ethnicity’ according to maternal countryof birth. Children were categorized as ‘Nordic’ if thematernal country of birth was Denmark, Finland,Iceland, Norway, or Sweden. We did not have accessto clinical data in the Finnish register. Patients fromFinland exctracted from the Nordic registers (D, I,N, S) were designated ‘Nordic’. Children to mothersoriginating from all other countries were definedas ‘non-Nordic’. In order to allocate non-Nordicchildren into suitable groups, the subjects werestratified according to five regions defined by a slightlymodified WHO administrative classification (22). InSwediabkids, data on the specified (maternal) countryof birth was available but blinded for the investigatorsbecause of governmental restrictions. Maternalcountry of birth could therefore only be included assubgroups according to Nordic/non-Nordic and theWHO subgroups of ethnicity.

Glycated HbA1c

Glycemic control was assessed by measurements ofHbA1c from capillary (D, I, S) and venous (N) bloodsamples. In Denmark and Norway, an annual HbA1cvalue from each patient is recorded in the registers. ThisHbA1c is centrally analyzed in reference laboratoriesat Herlev Hospital (Copenhagen) and Oslo UniversityHospital (Aker), respectively, using a high pressureliquid chromatographic method (Tosoh Bioscience®,

South San Francisco, CA, USA). The Diabetec Con-trol and Complications Trial (DCCT) HbA1c referenceinterval in Denmark is 4.3–5.8% corresponding to theInternational Federation of Clinical Chemistry (IFCC)HbA1c 23–40 mmol/mol and, in Norway, 4.6–6.0%,HbA1c (IFCC) 27–42 mmol/mol (23). The HbA1c(IFCC) mean and SD from target was +2.7 ± 1.9mmol/mol for Denmark (24). The Norwegian HbA1cdeviation from target was between −0.1 and 0.1%depending of the HbA1c level, SD was 1.6%, and thiswas irrespective of the IFCC/DCCT unit. The HbA1cvalues retrieved from both laboratories are validatedmonthly by the European Reference Laboratory andaligned with the DCCT/IFCC values. In Sweden,all HbA1c methods are standardized to the Mono Sprocedure. The constant relation between the MonoS procedure and the IFCC reference measurementprocedure is continuously validated in the IFCC moni-toring program. Each device participates in the Equalis(Equalis AB, Uppsala, Sweden) external quality assess-ment program. One fresh central sample is distributedten times a year for validation. The official referenceinterval for pediatric patients was HbA1c (Mono-S)4.6–5.0% [HbA1c (IFCC) 27–42 mmol/mol] (25). For2006–2009, the HbA1c mean and deviation from tar-get in Sweden was HbA1c (Mono S) −0.091 ± 0.044%,[HbA1c (IFCC) −0.9 ± 0.45 mmol/mol] (26). In Ice-land, HbA1c is measured with the DCA 2000 VantageAnalyzer (Siemens AG, Erlangen, Germany). TheHbA1c measurement is built on ‘monoclonal antibodyagglutination reaction’ with a reference interval of

Pediatric Diabetes 2014: 15: 519–527 521

Fredheim et al.

HbA1c (DCCT) 4.2–6.5%, 95% CI 4.3–5.7% [HbA1c(IFCC) 22–47 mmol/mol, 95% CI 23–39 mmol/mol].

Ethics approval

The study is approved by the respective national DataInspection Boards, public authorities for the protectionof individual privacy.

Statistical analyses

The grand means of continuous variables duringfollow-up were calculated based on the subject meanand followingly by calculating the grand mean ofthe population. Summary statistics are expressed asmeans ± SD. Logistic regression models for ordinalmeasures were applied to estimate p values fordemographical data. To determine the influence ofethnicity, we applied repeated measurement modelswith HbA1c as the outcome variable. We correctedfor differences in age at onset, diabetes duration, BMIz-score (continuous variables), and sex by includingthese explanatory variables in the model. Ethnicity(Nordic or non-Nordic) and country (D, I, N, andS) were considered class variables. Variables includedfor further analysis in the multivariate model weretreatment modalities [continuous subcutaneous insulininjection (CSII) or multiple daily insulin injection(MDI)], insulin requirement (U/kg/d), percentage ofdaily fast acting bolus insulin and the number of dailybolus doses. Several Danish patients were treated withinsulin glargine (generally administered once daily);whereas others were treated with pre-mixed insulin,NPH (neural protamin Hagedorn) and insulin detemir(which both are administered twice daily). In theDanish register (in 2006), only the total number ofinsulin injections was recorded from which the numberof doses with rapid acting insulin (bolus) was calculated(by means of subtraction). The predictors describedwere tested through forward elimination to evaluate theextent to which the differences between Nordic groupswere explained by treatment regimens. The variablesthat were significant predictors for HbA1c remained inthe multivariate model. A constant correlation withinthe individual patient over time was assumed for therepeated measures of HbA1c (some patients had severalHbA1c measurements while others had only a few).Statistical analyses were performed using sas version9.2 (SAS Institute, Inc. Cary, NC, USA). A p value<0.05 was considered statistically significant.

Results

The study population comprised 11,908 children andadolescents below age 16 yr at diagnosis and below18 yr at last visit (Fig. 1). Of these 6314 (53.0%) were

boys. A total of 921 (7.7%) patients originated fromnon-Nordic countries. The frequencies of non-Nordicsubjects were 5.7% (D), 2.7% (I), 5.5% (N), and 9.4%(S), respectively. A description of the non-Nordiccohort according to maternal birth region and studycountry is given in Fig. 2. The mean age (±SD) of thestudy cohort at diabetes onset was 7.7 (3.7) yr and theoverall mean duration of diabetes was 6.1 (3.6) yr.

The male:female ratio did not differ between Nordicand non-Nordic patients (p = 0.07, Table 1). Generally,patients with non-Nordic backgrounds were observedto have a lower mean (SD) age at diabetes onset [7.3(3.8) yr] compared to their native peers [7.8 (3.9) yr,p = 0.0008, unadjusted] and according to country; seeTable 1. This was not evident in Iceland, where thethree non-Nordic children had a higher age at onset,and the Icelandic cohort was generally older at diseaseonset compared to the other three countries. Durationof diabetes and BMI z-score did not differ betweenNordic vs. non-Nordic groups (Table 1).

Diabetes treatment characteristics

The most consistent difference between Nordic vs.non-Nordic groups was a more frequent use of CSIIin the Nordic patients (except for Norway) comparedto their non-Nordic peers (p < 0.0001, Table 1).Large discrepancies existed between countries: inDenmark and Sweden, the frequency of non-Nordicpatients using CSII were observed as 8 and 19%,respectively, whereas a majority of the non-Nordicpopulation (54%) from the Norwegian register usedCSII (p < 0.0001). Furthermore, non-Nordic childrendisplayed a slightly lower number of daily insulin bolusinjections compared to their Nordic peers (p < 0.0001).There was no difference in the percentage of bolusinsulin used between Nordic vs. non-Nordic groups(p = 0.5). However, Denmark was observed to have asignificantly lower percentage of daily bolus insulin usecompared to Norway and Sweden (p < 0.0001). Non-Nordic patients showed a higher HbA1c (p < 0.0001),although they had a slightly higher total insulin dosecompared to their Nordic counterparts (p = 0.001).

Glycemic control

The difference in HbA1c (IFCC) between Nordicand non-Nordic patients were significantly higher forthe non-Nordic patients in all four countries, alsowhen accounting for complex diabetes characteristics(Table 2). The HbA1c discrepancy between Nordic andnon-Nordic patients was lower for Sweden (0.6) thanfor Denmark and Norway (2.9 and 2.6, respectively)after adjusting for confounders (model 5, Table 2). TheHbA1c discrepancy was 1.2 with all countries com-bined accounting for all covariates (model 5, Table 2).



Tabl

e1.

Cha

ract

eris

tics

ofth

est

udy

coho

rt(1

1,90

8pa

tient

s).O

rdin

alva

riabl

esar

eex

pres

sed

asm

ean

(±S

D);

base

don

the

subj

ectm

ean

and

ther

eafte

rby

calc

ulat

ion

ofth

egr

and

(pop

ulat

ion)

mea

n

Var

iabl

eD

enm

ark

(D)2

322

Icel

and

(I)11

3N

orw

ay(N

)273

8S

wed

en(S

)673

5

Nor

dic*

Non

-Nor

dic†

Nor

dic

Non

-Nor

dic

Nor

dic

Non

-Nor

dic

Nor

dic

Non

-Nor

dic

pva

lue‡

pva

lue§

Pat

ient

s,n

(%)

2190

(94.

3)13

2(5

.7)

110

(97.

3)3

(2.7

)25

87(9

4.5)

151

(5.5

)61

00(9

0.6)

635

(9.4

)S

ex0.

070.

07M

ale

1144

5955

213

8468

3268

334

Fem

ale

1046

7355

112

0383

2832

301

M/F

ratio

1.09

0.80

1.00

0.50

1.15

0.83

1.15

1.11

Age

atdi

abet

eson

set;

year

s7.

99(3

.88)

7.40

(3.7

5)8.

80(3

.65)

9.45

(6.8

4)7.

77(3

.75)

6.70

(3.6

8)7.

67(3

.92)

7.45

(3.8

1)0.

001

0.00

17A

geat

late

stvi

sit;

yr12

.2(3

.5)

11.5

(3.8

)13

.5(3

.0)

16.3

(n/a

)12

.3(3

.5)

11.2

(3.7

)12

.5(3

.7)

12.0

(3.8

)<

0.00

01<

0.00

01D

urat

ion

ofdi

abet

es;y

r5.

83(3

.46)

5.80

(3.5

2)5.

78(3

.40)

4.06

(1.5

7)5.

79(3

.45)

5.61

(3.4

8)6.

28(3

.71)

5.96

(3.5

4)0.

020.

04B

MIz

-sco

re‖

0.3

(0.9

)0.

3(1

.3)

n/a

n/a

0.4

(1.1

)0.

5(1

.0)

0.4

(0.9

)0.

4(1

.0)

0.4

0.00

03In

sulin

regi

men

<0.

0001

<0.

0001

Mul

tiple

daily

insu

linth

erap

y;%

7492

9410

058

4669

81C

SII;

%26

86

042

5431

19D

aily

bolu

sin

ject

ions

2.4

(1.5

)1.

9(1

.2)

n/a

n/a

5.2

(1.3

)5.

0(1

.3)

4.4

(0.8

)4.

3(0

.8)

<0.

0001

<0.

0001

Bol

us(fa

st-a

ctin

g)in

sulin

;%of

tota

lda

ilydo

se39

.7(1

8.6)

33.0

(16.

2)n/

an/

a50

.7(1

7.3)

52.4

(17.

4)52

.0(1

2.5)

52.1

(13.

1)0.

5<

0.00

01

Insu

lindo

se;U

/kg/

d1.

0(0

.4)

1.1

(0.3

)0.

90(0

.3)

0.8

(-)¶

1.0

(0.4

)1.

0(0

.5)

0.9

(0.3

)1.

0(0

.5)

0.00

1<

0.00

01H

bA1c

(NG

SP

);%

8.3

(1.3

)8.

6(1

.3)

8.2

(1.4

)8.

1(−

)¶8.

4(1

.3)

8.7

(1.6

)8.

0(1

.3)

8.1

(1.5

)<

0.00

01<

0.00

01H

bA1c

(IFC

C);

mm

ol/m

ol67

.4(1

4.1)

70.3

(14.

6)65

.2(1

5.47

)65

.0(−

)¶68

.7(1

4.3)

71.5

(17.

6)63

.5(1

4.2)

64.8

(16.

5)<

0.00

01<

0.00

01

BM

I,bo

dym

ass

inde

x;H

bA1c

,gly

cate

dhe

mog

lobi

n,n/

a,da

tano

tava

ilabl

e;S

D,s

tand

ard

devi

atio

n.*N

ordi

c,m

ater

nalc

ount

ryof

birt

hfro

mN

ordi

cco

untr

ies

(Den

mar

k,Fi

nlan

d,Ic

elan

d,N

orw

ay,o

rS

wed

en).

†Non

-Nor

dic,

mat

erna

lcou

ntry

ofbi

rth

inan

yco

untr

you

tsid

eth

eN

ordi

cco

untr

ies.

‡pva

lue

acco

rdin

gto

diffe

renc

esbe

twee

nN

ordi

cvs

.non

-Nor

dic

patie

nts

(adj

uste

dfo

rco

untr

ies)

.§p

valu

eac

cord

ing

todi

ffere

nces

betw

een

the

coun

trie

s(a

djus

ted

for

Nor

dic

grou

ps).

¶O

nly

one

patie

ntw

ithda

ta.

‖BM

Iz-s

core

coul

dno

tbe

calc

ulat

edfo

rpa

tient

sle

ssth

antw

oyr

ofag

e(8

15ch

ildre

n).


Fredheim et al.

Iceland had only three participants with a non-Nordicbackground. The effect of ethnicity on glycemic con-trol in the multivariate model persisted when Icelandwas omitted from the analyses. In the countries (D,I, and N) were ethnic sub-groups (specific country oforigin) were available for analysis of HbA1c; higherHbA1c was largely attributed to patients originatingfrom MENA and sub-Saharan Africa (who accountedfor almost half and one-third of the non-Nordic popu-lation, respectively, Fig. 2). The percentage of childrenwith non-Nordic background based on child country ofbirth was lower (3%) than compared to maternal origin(8%). However, our main conclusions did not deviatefrom the reported findings when performing the analy-ses based on ethnic origin of the child; where the differ-ence in HbA1c (IFCC) (all countries combined) when

accounting for all covariates was slightly higher [2.89(0.66–1.32), p > 0.0001] compared to the HbA1c levelfor maternal origin [1.2 (0.4–2.0), model 5, Table 2].

Discussion

Despite a large heterogeneity (e.g., cultural, socioeco-nomic, religious etc.) within the non-Nordic patientgroup, our data demonstrate differences in treatmentregimens between Nordic and non-Nordic patients. Wefound a significantly slightly higher HbA1c among non-Nordic patients in all countries, also after adjustingfor diabetes treatment characteristics. Norway offeredCSII to all new patients; consequently, the use ofCSII was higher in Norway and particularly in chil-dren with non-Nordic background. The other countrieshad lower frequencies of CSII use overall – markedly

Fig. 2. Ethnic background of the study cohort stratified according to the World Health Organization. Participants from the Icelandic registerare not included in the figure (two participants with maternal country of birth from the Americas and one from Europe). Classificationof countries according to the World Health Organization: Europe: Albania, Armenia, Austria, Belgium, Bulgaria, Czech Republic, France,Germany, Greece, Hungary, Ireland, Italy, Luxemburg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Spain, Switzerland,Turkey, Ukraine, United Kingdom, Russia and other countries on European borders. Americas: Argentina, Bolivia, Brazil, Canada, Chile,Colombia, Costa Rica, Cuba, Ecuador, El Salvador, Guatemala, Haiti, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay,U.S.A. and Venezuela. MENA (Middle East and Northern Africa): Afghanistan, Algeria, Egypt, Iran, Iraq, Jordan, Lebanon, Libya,Mauritania, Morocco, Oman, Pakistan, Saudi Arabia, Sudan, Syria, Tunisia, United Arab Emirates and Yemen. Africa (Sub-Saharan): allAfrican countries except for Algeria, Egypt, Libya, Mauritania, Morocco, Sudan and Tunisia. Asia: Bangladesh, Bhutan, Cambodia, China,Fiji, India, Indonesia, Japan, Korea, Malaysia, Maldives, Mongolia, Myanmar, Nepal, Philippines, Sri Lanka, Thailand and Vietnam.

Table 2. Regression models to predict HbA1c between ethnic groups according to diabetes care (11,908 patients). TheHbA1c (IFCC, mmol/mol) regression coefficient [mean (CI)] listed for the non-Nordic patients. For example in Denmark,non-Nordic patients had a predicted HbA1c of 3.1 (0.6–5.6) mmol/mol higher than Nordic patients (base model). Whenaccounting for all covariates, the predicted HbA1c difference decreases to 2.9 (0.3–5.6) mmol/mol (model 5). The largeestimates for Iceland is due to scarce numbers

Statistical model n

Model HbA1c (regression coefficient)Denmark(D) 2322

Iceland(I) 113

Norway(N) 2738

Sweden(S) 6735

Allcountries

11908 p value† p value‡

1 Base model* 3.1 (0.6–5.6) 33.9 (13.5–54.3) 2.3 (0.4–4.9) 1.3 (0.3–2.4) 1.3 (0.3–2.4) 0.005 0.032 Base model + I 3.1 (0.7–5.6) 33.7 (1.8–22.9) 2.4 (0.1–4.9) 0.9 (0.1–1.9) 1.4 (0.6–2.2) 0.001 0.0023 Base model + I + R 2.4 (0–4.8) 33.2 (16.3–50.1) 2.4 (0.1–4.9) 1.0 (0.0–2.0) 1.4 (0.6–2.3) 0.0017 0.014 Base model + I + R + B 2.5 (0–4.8) n/a 2.5 (0–5.0) 0.9 (0.1–1.9) 1.4 (0.5–2.2) 0.0058 0.00455 Base model + I + R + B + N 2.9 (0.3–5.6) n/a 2.6 (0–5.2) 0.6 (0.4–1.6) 1.2 (0.4–2.0) 0.02 0.02

BMI, body mass index; CI, confidence interval; HbA1c, glycated hemoglobin; n/a, data not available.*Base model with HbA1c as the dependent variable adjusted for the following predictors: sex, age at onset, BMI z-score, diabetes durationand calendar year. Model terms are: I = insulin dose (U/kg/d), R = insulin regimen, B = % of fast-acting bolus injections of total insulin dose,n = Number of daily bolus injections.†p value according to all four countries combined.‡p value according to countries with the exclusion of Iceland (with three non-Nordic patients only).



lower in the non-Nordic children. The low frequencyand percentage of bolus insulin in Denmark was biasedby the more frequent use of pre-mixed insulin (count-ing as basal insulin injected only once to twice daily).Additionally, the implementation of CSII and morefrequent use of novel long-acting insulins, contributedto the observed change in recordings of bolus insulin.

Low incidence risk populations report a markedlyhigher age at disease onset than our cohort (27).We found that the non-Nordic cohort was youngerat disease onset compared to their native peers.These findings may indicate an accelerated diseaseprocess in non-Nordic patients, resulting in an earlierage at disease onset. BMI z-scores were somewhatsimilar between Nordic and non-Nordic patients,consequently, BMI did not provide an explanationfor the reported difference in age at disease onset. Weobserved a trend toward female excess in the non-Nordic group, which is typically seen in low incidencepopulations (6).

The discrepancy in glycemic control between Nordicgroups could to some extent be explained bytreatment regimen; although when all countries werecombined into one group, the difference in HbA1cwas diminutive, but persisted. The impact of varioustreatment centers could be argued to influence glycemiccontrol between Nordic groups, but our findings inHbA1c, however, were markedly smaller comparedto center-differences in HbA1c reported from theHvidoere study (28–30).

Non-Nordic patients had higher HbA1c in allcountries – even in Norway where insulin pumpswere more frequently used in non-Nordic patients.Adherence to the prescribed treatment is pivotal forglycemic control. The use of CSII therapy requires abetter understanding of diabetes and relies on a highstandard of self-management compared to insulin pentreatment; it includes more daily self-testing of glucose,dietary changes (including carbohydrate counting),extra adjustments of prandial insulin, and physicalexercise, among other things. This has improvedconsiderably on the hands of patients and professionalsthe recent years. This may explain the diminutivedifference in glycemic control between Nordic groupsreported here, alongside equal opportunities fordiabetes care.

On the other hand, the use of CSII is not inevitablyaccompanied with reductions in HbA1c, but ishighly dependent upon adherence to the prescribedtreatment. The criteria for initiating CSII therapydiffered between the four countries and have changedremarkably during the years from family requeststo high HbA1c, and recurrent severe hypoglycemia.Insulin pumps were first introduced in Denmark in2005, and their use increased to approximately 30%of the type 1 diabetes population in 2009 (data not

shown). In Sweden, CSII use increased from 24 to29% during the same period (31). Through physicianevaluation of the pre-requisites for pump treatment,some patients do not qualify for insulin pump use,possibly because of language difficulties, poor com-pliance, or numeracy/literacy. In Norway, CSII wasimplemented approximately 6 yr earlier than Denmarkand Sweden, and they primarily administered CSIIto patients with high HbA1c, very young patients orpatients with recurrent severe hypoglycemia. Since2000, Oslo University Hospital, with the majority ofthe non-Nordic population in the country, offeredinsulin pumps to all patients from the onset of diabetes.Consequently, Norway presented with the highest per-centage of children on insulin pumps and with a higherpercentage of pump users in the non-Nordic children.Currently, other pediatric centers in Norway also offerinsulin pumps to the majority of their patients.

Studies from the US report higher mean HbA1clevels in African–American youths +1.5 and +0.4%(32, 33), respectively, compared to Caucasians orNon-Hispanic whites. Also an Australian study foundhigher HbA1c in Maori (+1.0%) and Pacific (+1.2%)patients compared to patients from Europe (16). Previ-ously in Denmark, ethnic minorities were reported with+0.4% (DCCT) unit higher HbA1c levels than theirnative Danish peers (34). These results reflect largerdifferences in HbA1c between ethnic groups than wefound between Nordic groups. The studies have con-siderably fewer (146–555) patients than our study; butethnicity was also defined as either maternal ethnicityor child origin (data extracted from medical charts,self-identified, or through patient questionnaires). Theheterogeneity of the (national) background populationis reflected in data analyses, as ethnic sub-groups arenot similarly defined across the countries, however,there is a consistency in the results where the respectivenative populations had significantly lower Hb1c levelsthan minority groups. There is marked differencesbetween countries in health care services, which inthe US primarily is based on insurance coverage,and is not tax financed as in the Nordic countries.Access to health care offer a plausible explanation tothe reported difference in glycemic control betweenethnic groups across countries; free health care maycontribute to a reduction in this discrepancy.

We stratified our patients according to maternalcountry of birth on the assumption that maternalethnicity influenced the level of diabetes care for theparticipating child. With this partition, non-Nordicpatients accounted for approximately 8% of the cohortoverall. As expected, ethnicity based on birth countryof the child was lower (3%) for the non-Nordic group.It would have been valuable to investigate paternalethnicity, but unfortunately, because of governmentalrestrictions and too many missing data for this


Fredheim et al.

variable, this was not applicable. The overall resultsby use of maternal country of birth as a proxy forethnicity were unambiguous in this study cohort,and comprise a valuable parameter to investigatedifferences in treatment paradigms and outcomebetween minority groups.

It is a challenge to compare results across borders,and the composition of the populations varies acrossthe Nordic countries. National standards for healthcare, education, and training programs constituteconfounding factors that are equally important as freehealth care, and may offer additional explanationsto the reported discrepancy in glycemic control.Altogether, health care services are comparable withinthe Nordic countries, and provide equal access andno direct economic hindrance in seeking medicalcare. Targeted intervention for integrated care indifferent cultural settings is a cornerstone for successfulintervention and must be prioritized.

Strengths and weaknesses

Data comprise a large study cohort providing a uniqueopportunity to analyze confounders affecting glycemiccontrol across four population-based registries with ahigh data ascertainment. The impact of diabetes carecould be quantified, although some differences betweencountries existed, with the results pointing in thesame direction irrespective of country. Methodologyhas to be taken into account when comparingHbA1c between countries. HbA1c measurements wereanalyzed centrally or standardized on a national level,thereby reducing methodological bias. We specified thedeviations from target but chose to present unadjustedHbA1c because they are validated domestically. Webelieve that the HbA1c values are reliable, althoughsome would argue that central measurements ofHbA1c across national standards should be obtainedto optimize data validity.

Our findings were limited by disparities in datadesign between the diabetes registers. Data unification,governmental restrictions and data protection pre-vented us from delineating on SES and intra-familialdifferences with an effect on HbA1c. This studyhighlights the challenges emerging when data fromdifferent registration methods are combined and callsfor collaboration across countries in order to improvedata unification and interpretation.

An attempt to pinpoint only a few predictors forglycemic control in ethnic groups may result in askewed picture of reality. Cultural, socioeconomic,and religious aspects must be considered when ethnicdifferences are investigated. However, the differencesbetween Nordic and non-Nordic patients reportedhere are less pronounced than differences betweenethnic groups previously reported, and we believe that

equal access to diabetes care contributes significantlyto this uplifting finding. Disparities in glycemiccontrol, although diminutive, highlights the need fortargeted intervention in high risk groups, irrespectiveof ethnic background.

Acknowledgements

We are very grateful to the patients and their families andthe pediatricians and nurses who have provided us withpatient information. Special thanks to Birthe Olsen (D), AndersJohansen (D), Niels Birkebæk (D), Jesper Johannesen (D) fromDanDiabKids, Knut-Dahl Jørgensen (N), Geir Joner (N) fromNCDR, Ulf Samuelsson (S), Lena Hanberger (S) and KarinAkesson (S) from Swediabkids. Jannet Svensson is guarantorfor these data. This work was supported by the Nordic DiabetesRegistries (Denmark, Norway), the Icelandic Thorvaldsen’sFoundation (Island) and the Swedish Board of Health andWelfare (Swediabkids).

Conflict of interest

The authors declare no conflict of interest.

References

1. Patterson CC, Dahlquist GG, Gyurus E, Green

A, Soltesz G, EURODIAB Study Group. Incidencetrends for childhood type 1 diabetes in Europeduring 1989-2003 and predicted new cases 2005-20: amulticentre prospective registration study. Lancet 2009:373: 2027–2033.

2. Bell RA, Mayer-Davis EJ, Beyer JW et al. Diabetesin non-Hispanic white youth: prevalence, incidence,and clinical characteristics: the SEARCH for Diabetesin Youth Study. Diabetes Care 2009: 32 (Suppl 2):S102–S111.

3. Harjutsalo V, Sjoberg L, Tuomilehto J. Time trendsin the incidence of type 1 diabetes in Finnish children: acohort study. Lancet 2008: 371: 1777–1782.

4. Berhan Y, Waernbaum I, Lind T, Mollsten A,Dahlquist G. Swedish Childhood Diabetes StudyGroup. Thirty years of prospective nationwide incidenceof childhood type 1 diabetes: the accelerating increaseby time tends to level off in Sweden. Diabetes 2011: 60:577–581.

5. DIAMOND Project Group. Incidence and trends ofchildhood type 1 diabetes worldwide 1990-1999. DiabetMed 2006: 23: 857–866.

6. Tuomilehto J. The emerging global epidemic of type 1diabetes. Curr Diab Rep 2013: 13: 795–804.

7. Delli AJ, Lindblad B, Carlsson A et al. Type 1diabetes patients born to immigrants to Sweden increasetheir native diabetes risk and differ from Swedishpatients in HLA types and islet autoantibodies. PediatrDiabetes 2010: 11: 513–520.

8. Bodansky HJ, Staines A, Stephenson C, Haigh D,Cartwright R. Evidence for an environmental effectin the aetiology of insulin dependent diabetes in atransmigratory population. BMJ 1992: 304: 1020–1022.

9. Ehehalt S, Popovic P, Muntoni S et al. Incidence ofdiabetes mellitus among children of Italian migrants



substantiates the role of genetic factors in thepathogenesis of type 1 diabetes. Eur J Pediatr 2009:168: 613–617.

10. Hjern A, Soderstrom U, Aman J. East Africans inSweden have a high risk for type 1 diabetes. DiabetesCare 2012: 35: 597–598.

11. Kirk JK, Passmore LV, Bell RA et al. Disparities inA1C levels between Hispanic and non-Hispanic whiteadults with diabetes: a meta-analysis. Diabetes Care2008: 31: 240–246.

12. Svensson M, Eriksson JW, Dahlquist G. Earlyglycemic control, age at onset, and development ofmicrovascular complications in childhood-onset type 1diabetes: a population-based study in northern Sweden.Diabetes Care 2004: 27: 955–962.

13. Skrivarhaug T, Bangstad H-J, Stene LC, Sandvik L,Hanssen KF, Joner G. Low risk of overt nephropathyafter 24 yr of childhood-onset type 1 diabetesmellitus (T1DM) in Norway. Pediatr Diabetes 2006: 7:239–246.

14. Skrivarhaug T, Fosmark DS, Stene LC et al. Lowcumulative incidence of proliferative retinopathy inchildhood-onset type 1 diabetes: a 24-year follow-upstudy. Diabetologia 2006: 49: 2281–2290.

15. Gallegos-Macias AR, Macias SR, Kaufman E,Skipper B, Kalishman N. Relationship betweenglycemic control, ethnicity and socioeconomic statusin Hispanic and white non-Hispanic youths with type 1diabetes mellitus. Pediatr Diabetes 2003: 4: 19–23.

16. Carter PJ, Cutfield WS, Hofman PL et al.Ethnicity and social deprivation independently influencemetabolic control in children with type 1 diabetes.Diabetologia 2008: 51: 1835–1842.

17. Tubiana-Rufi N, Moret L, Czernichow P, Chwalow

J. Risk factors for poor glycemic control in diabeticchildren in France. Diabetes Care 1995: 18: 1479–1482.

18. Svensson J, Marinelli K, Eising S. Danish Child-hood Diabetes Register, National Discharge Register.[Comparison of registration of data from the DanishChildhood Diabetes Register and The NationalDischarge Register]. Ugeskr Laeger 2007: 169: 122–125.

19. Skrivarhaug T, Stene LC, Drivvoll AK, Strøm H,Joner G. Norwegian Childhood Diabetes Study Group.Incidence of type 1 diabetes in Norway among childrenaged 0-14 years between 1989 and 2012: has the incidencestopped rising? Results from the Norwegian ChildhoodDiabetes Registry. Diabetologia 2014: 57: 57–62.

20. WHO. Definition and Diagnosis of Diabetes Mellitusand Intermediate Hyperglycaemia [Internet]. WHO2013 (available from http://www.who.int/diabetes/publications/diagnosis_diabetes2006/en/index.html)

21. Kuczmarski RJ, Ogden CL, Grummer-Strawn LMet al. CDC growth charts: United States. Adv Data2000: 314: 1–27.

22. WHO. WHO Regional Offices [Internet]. WHO2013 (available from http://www.who.int/about/regions/en/index.html)

23. National Glycohemoglobin Standardization Program(NGSP) 2014 (available from http://www.ngsp.org/convert1.asp).

24. Klinisk Biokemi i Norden; Bias av HbA1c upptacktmed EQA i Skandinavien. Bias Av HbA1c UpptacktMed EQA Skand. Nr. 2, vol. 25, 2013 (available fromhttp://kbn.nfkk.org/kbn_2013_2/)

25. Swedish Society for Clinical Chemistry. 2014 (availablefrom http://www3.svls.se/sektioner/sfkk/links/qastand.htm).

26. Lindblad B, Nordin G. External quality assessment ofHbA1c and its effect on comparison between Swedishpediatric diabetes clinics. Experiences from the Swedishpediatric diabetes quality register (Swediabkids) andEqualis. Clin Chem Lab Med 2013: 51: 2045–2052.

27. Blumenfeld O, Dichtiar R, Shohat T, The IsraelIDDM Registry Study Group (IIRSG). Trends in theincidence of type 1 diabetes among Jews and Arabs inIsrael. Pediatr Diabetes 2013. doi: 10.1111/pedi.12101[Epub ahead of print].

28. Danne T, Mortensen HB, Hougaard P et al.Persistent differences among centers over 3 years inglycemic control and hypoglycemia in a study of 3,805children and adolescents with type 1 diabetes fromthe Hvidøre Study Group. Diabetes Care 2001: 24:1342–1347.

29. De Beaufort CE, Swift PGF, Skinner CT et al.Continuing stability of center differences in pediatricdiabetes care: do advances in diabetes treatmentimprove outcome? The Hvidoere Study Groupon Childhood Diabetes. Diabetes Care 2007: 30:2245–2250.

30. De Beaufort CE, Lange K, Swift PG et al. Metabolicoutcomes in young children with type 1 diabetes differbetween treatment centers: the Hvidoere Study in YoungChildren 2009. Pediatr Diabetes 2013: 14: 422–428.

31. Swediabkids. 2014 (available from https://swediabkids.ndr.nu/).

32. Auslander WF, Thompson S, Dreitzer D, White

NH, Santiago JV. Disparity in glycemic control andadherence between African-American and Caucasianyouths with diabetes. Family and community contexts.Diabetes Care 1997: 20: 1569–1575.

33. Kamps JL, Hempe JM, Chalew SA. Racial disparity inA1C independent of mean blood glucose in children withtype 1 diabetes. Diabetes Care 2010: 33: 1025–1027.

34. Povlsen L, Olsen B, Ladelund S. Diabetes in childrenand adolescents from ethnic minorities: barriers toeducation, treatment and good metabolic control. JAdv Nurs 2005: 50: 576–582.


equal access to health care may diminish the differences in outcome between native and immigrant...

Documents