Content uploaded by Jaakko Tuomilehto
Author content
All content in this area was uploaded by Jaakko Tuomilehto on Sep 23, 2014
Content may be subject to copyright.
The aetiology and natural history of Type I (insulin-
dependent) diabetes mellitus are still not known but
both genetics and environmental factors contribute
to the development of the disease [1±3]. Although
HLA genetics have a major role in the aetiology of
Type I diabetes, other genes also contribute to the ge-
netic effect, but the mode of inheritance of the dis-
ease is not clear [4]. The genetic effect contributes
70±75% of the susceptibility to Type I diabetes [5,
6]. Environmental factors possibly initiate or trigger
the process which leads to the destruction of the
beta cells and the onset of diabetes [3, 7, 8].
In the late 1970s epidemiological reports of dia-
betic children for the first time showed a wide geo-
graphical variation in the incidence of Type I diabe-
tes. During the 1960s to the early 1980s the data on
incidence of Type I diabetes were available for a few
populations only, mostly from regions with a high or
intermediate risk for this disease. A large number of
registries had been established since the mid 1980s
worldwide. The lack of standardized data made it dif-
ficult to determine the true magnitude of the world-
wide variation in incidence or time trends [9]. The Di-
Diabetologia (1999) 42: 1395±1403
Articles
Worldwide increase in incidence of Type I diabetes ±
the analysis of the data on published incidence trends
P. Onkamo, S. Väänänen, M. Karvonen, J. Tuomilehto
Diabetes and Genetic Epidemiology Unit, Department of Epidemiology and Health Promotion, National Public Health Institute,
Helsinki, Finland
Ó Springer-Verlag 1999
Abstract
Aims/hypothesis. Several reports on the incidence of
Type I (insulin-dependent) diabetes mellitus have
suggested that the incidence is increasing. The aim
of this study was to find out whether the incidence is
increasing globally or restricted to a selected popula-
tions only and to estimate the magnitude of the
change in incidence.
Methods. During 1960 to 1996 37 studies in 27 coun-
tries were carried out. To fulfil the inclusion criteria
the study periods ranged from 8±32 years. The tem-
poral trend was fitted by linear regression, with the
logarithm of the age-standardized incidence as the
dependent variable and the calendar year as the inde-
pendent variable. Then, the regression coefficient
( ´ 100%) is approximately the average relative in-
crease in incidence per year (as percentage).
Results. Results from the pooled data from all 37 pop-
ulations showed that the overall increase in incidence
was 3.0% per year (95% CI 2.6; 3.3, p = 0.0001). The
statistically significant increase was found in 24 of 37
populations including all high incidence ( > 14.6 per
100 000 a year) populations. The relative increase
was, however, steeper in the populations with a lower
incidence. The correlation between logarithm of the
incidence and the increase in incidence was
r = ±0.56, p = 0.0004.
Conclusion/interpretation. The incidence of Type I di-
abetes is increasing worldwide both in low and high
incidence populations. By the year 2010 the incidence
will be 50 per 100 000 a year in Finland and also in
many other populations it will exceed 30 per 100 000
a year. [Diabetologia (1999) 42: 1395±1403]
Keywords Epidemiology, geographical variation, in-
cidence, increase, modelling, prediction, trend, Type
I diabetes.
Received: 3 May 1999 and in final revised form: 17 August
1999
Corresponding author: J. Tuomilehto, Diabetes and Genetic
Epidemiology Unit, Department of Epidemiology and Health
Promotion, National Public Health Institute, Mannerheimin-
tie 166, FIN-00300 Helsinki, Finland
abetes Epidemiology Research International Group
(DERI) started the collection of the aggregate data
on incidence of Type I diabetes in the late 1970s and
the early 1980s [10]. The efforts of the DERI group
led to an increase in the number of registries on dia-
betic children and to the establishment of the World
Health Organization Project of Childhood Diabetes
(DIAbetes MONdiale) in 1990 [11]. The collabora-
tive research project EURODIAB ACE was estab-
lished also in the late 1980s [12] to gather information
of Type I diabetes in Europe. The latest reviews on
the incidence of Type I diabetes among populations
have indicated that differences in the incidence are
60-fold between the highest and the lowest rates [10,
12±14]. The highest incidence is found in Caucasoid
populations, particularly in northern Europe, and
the lowest rates are found in Asia and South America
[13, 14].
Thus far only one trend analysis of the incidence of
Type I diabetes comparing simultaneously several
but yet a limited number of populations has been car-
ried out by the DERI group [15]. The standardized
procedures agreed upon for the incidence data col-
lected around the world now permit a comparative
assessment of temporal trends among several popula-
tions. We estimated the temporal trends in the inci-
dence of Type I diabetes from incidence data collect-
ed through a systematic literature review. A statistical
analysis of the data was done in order to find out
whether the incidence is increasing globally. Another
objective was to evaluate quantitatively the extent to
which the change in incidence of Type I diabetes dif-
fers among populations.
Materials and methods
Literature search. The literature was searched using MED-
LINE, direct examination of reference lists of the articles,
hand searches of selected journals and published conference
abstracts. By the closing date, 28th February 1999, more than
160 original publications reporting time series of the incidence
of Type I diabetes were found.
Inclusion criteria. The publications were further evaluated
with the strict inclusion criteria in order to choose appropri-
ate studies for the quantitative analysis. The inclusion criteria
were: 1) the study period was 8 years or more, 2) the inci-
dence rates were presented for each year separately, 3) the
number of cases per year was five or more, 4) in the papers
in which the age standardization had been reported, the inci-
dences had been estimated with age standardization accord-
ing to the world population and, 5) Type I diabetes was diag-
nosed according to the WHO definition. Studies comparing
incidence rates estimated with different methods during dif-
ferent periods were excluded. No requirement for a minimum
case-ascertainment level of the data source was made as reli-
able case ascertainment estimation had usually not been
done until the 1990 s. The articles were evaluated by two inde-
pendent reviewers according to the above-mentioned inclu-
sion criteria.
Description of the data. Incidence data were obtained either
from the tables or from the figures in the published articles.
In approximately half of the original articles the annual inci-
dence rates were presented in the tables and in another half in
the figures only. The numbers derived from the figures were
reconfirmed by the second reviewer. For Montreal (Canada),
Allegheny County (USA), Scotland, Auckland (New Zea-
land), Prince Edward Island (Canada), Leicestershire (UK)
and Wielkopolska (Poland) we used the original database of
the DERI Study [15].
Altogether 37 studies from 27 countries met the inclusion
criteria and were included in the analysis (Table 1). The regis-
tration of diabetic children was prospective in most of the
studies. In 30 studies the age of children ranged from 0 to
14 years and in 7 studies from 0 to 15, 16, 17 and 19 years.
The time period of the studies ranged from 8 to 32 years.
The average length of the study period was 14.9 years (medi-
an 14 years). The estimates of the degree of case-ascertain-
ment were high, ranging from 85 to 100%. The degree of as-
certainment remained unspecified only in five studies. The
studies included in the analysis were from the period 1960 to
1996.
Statistical methods. The incidence of Type I diabetes for our
analysis was taken from the individual studies as it was report-
ed in these publications. The incidence for the data obtained
from the DERI Study [15] was calculated per 100 000 people
a year. Age standardization of the rates was done using 5-year
intervals with the proportions 33/100, 33/100 and 34/100 (for
the age groups 0±4 years, 5±9 years, and 10±14 years respec-
tively) as the standard according to the approach by the
DERI Study Group [16], which is the same as the world popu-
lation standard.
The temporal trend for each population was fitted by a sim-
ple linear regression under the assumption of normally distrib-
uted errors, with the age-standardized incidence as dependent
variable expressed on a logarithmic scale and the calendar
year as independent variable: lnl
i
(t) = a
i
+ b
i
t, where l
i
(t) de-
note the age-standardized incidence predicted at year t for
population i; the intercept a
i
is different for each population,
and b
i
is the population specific regression coefficient (the
trend), respectively. In such a multiplicative model the regres-
sion coefficient ( ´ 100 %) is a percentage, approximately be-
ing the average relative increase in incidence per year. The
multiplicative regression model was used because it fitted the
data well. It is commonly used in estimating time trends in inci-
dence and allows a simple interpretation of the regression co-
efficient.
The overall estimate of the relative annual increase was ob-
tained by using a pooled, centralized data set: to start, for each
population the logarithms of the age-standardized incidence
rates and the time points were centralized to make the differ-
ent lengths of the studies and incidence levels more compara-
ble. Then, using the method of least squares, a straight line
constrained to cross the origin of the centralized coordinate
system was fitted to the pooled data set. The regression coeffi-
cient has the same interpretation as in the population-wise
analysis. The analysis was subsequently repeated as weighted
regression, where the residual sum of squares was weighted
with the number of cases in individual studies, to give more
weight for observations with a higher number of cases.
The association between the level of incidence and increase
in incidence was assessed by calculating the correlation coeffi-
cient between the logarithms of incidences in 1983 predicted
by the model and the incidence increases estimated by the mul-
tiplicative model. The year 1983 was chosen because almost all
studies covered it.
P. Onkamo et al.: Worldwide increase in incidence of Type I diabetes1396
Predictions until the year 2010 have been made with both
multiplicative and additive regression models, meaning that
the curve produced by fitting the model to the data is simply
extrapolated to the year 2010. In essence, the multiplicative
model fits an exponential curve to the incidence, whereas the
additive model fits a straight line. The additive regression
model is used here to point out the differences between predic-
tions when using alternative models.
Results
Incidence. The mean incidence of Type I diabetes
among the study populations varied from 0.5 to 30.3
per 100 000 a year during the observation period (Ta-
ble 1). The mean incidences were divided into quar-
tiles: low incidence, less than 6.4 per 100 000 a year
(the lowest 25% of the mean incidences), intermedi-
ate 6.4±14.6 per 100 000 a year (50% of the mean in-
cidences, thus the two intermediate groups com-
bined), and high, more than 14.6 per 100 000 a year
(highest 25% of the mean incidences).
Increase in the incidence of Type I diabetes. The rela-
tive change (% per year) in incidence among individu-
al populations ranged from ±0.2% in Colorado (USA)
to 9.5% in Leicestershire (UK) (Table 2). A statisti-
cally significant increase in incidence was found in
65% (24/37) of the populations. An upward tendency
in incidence not reaching statistical significance was
observed in another 12 populations. Only in one pop-
ulation, Colorado (USA), the trend was slightly, but
not significantly, negative (±0.2% per year) and the
upper limit of the 95% confidence interval show that
an increase of 2.2% per year was possible.
The global trend and the annual increase in the in-
cidence of Type I diabetes were estimated from the
P. Onkamo et al.: Worldwide increase in incidence of Type I diabetes 1397
Table 1. The studies included in the analysis
Country: area Study
period
Length
of study
period
(years)
Age-
group
(years)
Number
of cases
Mean
incidence
(100,000/
year)
Study
design
a)
Degree of
ascertain-
ment
Rates
obtained
from
b)
Reference
Algeria: Oran 1979±1988 10 0±14 173
c)
4.7 P, R ? F 36
Australia: West 1985±1992 8 0±14 350 14.9 P 99.6% T 37
Austria 1979±1993 15 0±14 1551 7.8 P, R > 90% T 15, 38
Bulgaria, East 1974±1995 22 0±14 818 6.3 R, P 98.8% T 39
Canada: Montreal 1971±1985 15 0±14 839 9.3 R 94% T 40
Canada: Prince 1975±1986 19 0±14 115 23.5 P 99% T 41, 42
Edward Island 1990±1993
China: Shanghai 1980±1993 14 0±14 67 0.7 R 85.2% F 43, 44
East-Germany 1960±1989 30 0±19 9581 6.7 P 99% F 45
Estonia 1983±1996 14 0±14 523 10.2 P 92%
d)
96%
e)
T 42, 46
Finland 1965±1996 32 0±14 9047 30.3 P, R 100% T 14, 47
France 1988±1995 8 0±19 1439 8.0 P 96% T 48
Hungary 1978±1987 10 0±14 1060 6.1 P 96.2% T 49
Iceland 1970±1988 19 0±14 120 9.0 P, R 100% F 50
Israel: Yemenite Jews 1965±1993 29 0±17 1665 5.0 P ? F 51
Italy: Turin 1984±1991 8 0±14 227 8.4 R, P 99% F 52
Japan: Hokkaido 1973±1992 20 0±14 396 1.7 P 100% F 53
Latvia 1983±1996 10 0±14 652 7.2 P 80±100% F, T 54, 55
Libya 1981±1990 10 0±14 251 8.7 P ? T 56
Lithuania 1983±1992 10 0±14 215 6.4 P 95±100% F 54
Malta 1980±1996 17 0±14 90 14.7 P ? F 57
New Zealand: Auckland 1977±1987 11 0±15 206 10.1 P 100% T 58
New Zealand: Canterbury 1982±1990 9 0±19 123 12.7 P 100% F 59
Norway 1973±1982 10 0±14 1914 20.8 R 99.4% F 60
Peru: Lima 1985±1994 10 0±14 111 0.5 P 85 % T 61
Poland: Krakow 1987±1994 8 0±14 396 5.9 P 100% T 62
Poland: Rzeszów 1980±1992 13 0±14 122 5.1 P, R 99% F 63
Poland: Wielkopolska 1970±1985 15 0±16 451 4.4 P > 95% T 15, 64
Slovakia 1985±1995 11 0±14 1127 7.5 P 95% T, F 65, 66
Sweden 1978±1992 15 0±14 5831 24.9 P, R 100 % F 3
UK: Leicestershire 1965±1981 17 0±14 248 7.8 P, R 97.6% F 15
UK: Oxford 1985±1995 11 0±14 1037 18.5 P ? T 67
UK: Plymouth 1975±1996 22 0±14 488 14.9 P 95.3% F 68
UK: Scotland 1976±1993 18 0±14 4182 21.6 R > 95% T 15
UK: Yorkshire 1978±1992 15 0±14 1721 14.3 P 97.6% F 69, 70
USA: Allegheny County 1965±1985 21 0±14 1041 14.7 P, R 100 % T 71
USA: Colorado 1978±1988 11 0±17 1376 7.8 P, R 93.3% F 72
USA: Hawaii 1980±1990 11 0±14 113 12.3 P 97% F 73
a)
P = prospective, R = retrospective
b)
F = figure, T = table
c)
families
d
non Estonian
e
Estonian
age-standardized incidence rates using the log-linear
regression (Table 2). The p value of less than 0.05 for
the two-sided test for a non-zero regression coeffi-
cient was regarded as evidence for the trend. The
global annual increase was 3.0% (95% CI 2.59; 3.33,
p = 0.0001) during 1960 to 1996, showing a highly sig-
nificant increasing trend. When the annual incidence
rates were weighted with the number of cases in
each individual study, the increase in incidence was
2.5% (95% CI 2.32; 2.66; p = 0.0001). The estimated
population-wise regression lines illustrate well the in-
creasing trends (Fig.1, 2).
Comparison of increase rates. The relation between
the increase in incidence of Type I diabetes and the
average level of incidence expressed as the incidence
in 1983 estimated from the regression model is shown
in Figure 3. There was a significant inverse associa-
tion between the increase and the logarithm of the
level of incidence (r = ±0.56, p = 0.0004). The associa-
tion indicates that the relative increase was more pro-
nounced in the populations with a low incidence.
Nevertheless, in the five populations with the highest
incidence the increase was also statistically signifi-
cant, varying from 1.2 to 3.2% per year.
The incidence level of Type I diabetes and its in-
crease seemed to be similar in some geographically
adjacent populations. For example in the northern
European countries; Finland, Sweden and Norway
where the incidence of Type I diabetes has been high
for a long time, the increase was 1 to 3% per year.
Adjacent countries around the Baltic Sea, Estonia,
Latvia, Lithuania and Poland with an intermediate
or low incidence (4±10 per 100 000 a year) showed
P. Onkamo et al.: Worldwide increase in incidence of Type I diabetes1398
Table 2. Relative increase in incidence of Type I diabetes in
children aged 14 years or less in 37 populations. The popula-
tions are arranged in descending order according to the rela-
tive increase per year. The 95% confidence interval is given in
the parentheses. P value stands for the two-sided test for a non-
zero regression coefficient
Country: area Mean incidence Increase in incidence %
per year (95% CI)
p value
UK: Leicestershire 7.8 9.5 (6.51; 12.53) 0.0001
Hungary 6.1 8.5 (6.50; 10.42) 0.0001
Algeria: Oran 4.7 7.9 (1.85; 14.00) 0.0338
USA: Hawaii 7.8 7.8 (1.80; 14.87) 0.0315
Peru: Lima 0.5 7.7 (0.97; 16.40) 0.1197
China: Shanghai 0.7 7.4 (2.30; 12.51) 0.0148
Poland: Krakow 5.9 6.8 (2.27; 11.41) 0.0262
New Zealand: Auckland 10.1 6.4 (4.20; 8.52) 0.0003
Australia: West 14.9 6.3 (2.11; 10.50) 0.0259
Libya 8.7 6.3 (0.69; 11.8) 0.0589
Japan: Hokkaido 1.7 5.9 (4.14; 7.63) 0.0001
Slovakia 7.5 5.5 (3.64; 7.41) 0.0003
Poland: Wielkopolska 4.4 4.8 (1.94; 7.66) 0.0054
France 8.0 3.9 (2.85; 4.94) 0.0003
UK: Oxford 18.5 3.7 (1.82; 5.50) 0.0037
Canada: Prince Edward Island 23.5 3.2 (0.33; 6.38) 0.0728
Israel: Yemenite Jews 5.0 3.2 (2.51; 3.88) 0.0001
Norway 20.8 3.2 (1.19; 5.22) 0.0143
Austria 7.8 2.7 (1.58; 3.76) 0.0003
UK: Plymouth 14.6 2.7 (0.91; 4.50) 0.0079
New Zealand: Canterbury 12.7 2.7 (0.05; 10.50) 0.5262
UK: Scotland 21.6 2.5 (1.85; 3.08) 0.0001
East-Germany 6.7 2.4 (1.96; 2.90) 0.0001
Finland 30.3 2.3 (1.98; 2.57) 0.0001
Iceland 9.0 2.3 (2.38; 6.96) 0.3498
Italy: Turin 8.4 2.2 (3.99; 8.35) 0.5150
Bulgaria: East 6.3 2.1 (1.03; 3.15) 0.0010
UK: Yorkshire 14.3 1.9 (0.30; 3.53) 0.0372
Canada: Montreal 9.3 1.6 (0.67; 3.82) 0.1937
USA: Allegheny County 14.7 1.5 (0.21; 2.83) 0.0348
Sweden 24.9 1.2 (0.42; 2.02) 0.0103
Poland: Rzeszów 5.1 1.1 (3.25; 5.40) 0.6339
Lithuania 6.4 1.1 (4.25; 6.41) 0.7023
Latvia 7.2 0.9 (1.90; 3.75) 0.5350
Malta 14.7 0.5 (2.15; 3.19) 0.7078
Estonia 10.2 0.4 (0.96; 1.76) 0.5741
USA: Colorado 12.3 0.2 (2.52; 2.19) 0.8938
Change globally 3.0 (2.59; 3.33) 0.0001
Change globally (weighted by number of cases) 2.5 (2.32; 2.66) 0.0001
an upward course but not a statistically significant
trend in incidence. The increase in incidence in east-
ern Europe varied from 2.1% per year in East Bul-
garia to 8.5% per year in Hungary. In the United
Kingdom the mean incidence ranged from 14.3 to
21.6 and the increase in incidence ranged between
1.9 and 3.7% except for Leicestershire where the
mean incidence was 7.8 with an increase of 9.5%.
The data for Leicestershire, however, were consider-
ably older (from 1965 to 1981) than from other UK
study populations.
Predictions until the year 2010. Since no effective pre-
vention has thus far been invented or is foreseen in
the near future, we used the observed trends to pre-
dict the incidence of Type I diabetes at least until the
year 2010 (Table 3). For the prediction we applied
both linear and exponential models, since the model
of the increase in incidence is not known. The expo-
nential predictions were only calculated for popula-
tions with a study period of at least 14 years. In gener-
al, the linear model produces more conservative pre-
dictions than the exponential model, however, in
those populations where an increase had started dur-
ing the very last years of the observation period, the
exponential model gave more conservative predic-
tions than the linear.
The predictions based on the linear model show
that Finland will still have the highest incidence in
the world (50 per 100 000 a year) in the year 2010, fol-
lowed by Norway, Prince Edward Island (Canada),
western Australia, Scotland (UK), Oxford (UK),
and Sweden. Despite the large relative increases in
the incidence observed in China and Peru, the abso-
lute incidence rates in these countries would still re-
main low, less than 2 per 100 000 a year. Based on
these predictions, in Japan the incidence will be lower
than 5 per 100 000 a year and in Poland, Latvia and
Lithuania the incidence will be under 10 per 100 000
a year.
Discussion
There are also other populations in which an increase
in the incidence of Type I diabetes has been recently
reported such as Croatia, Denmark, Kuwait, the
Netherlands, Russia and Switzerland [17±23]. These
studies, however, did not meet the inclusion criteria
of this study and were not included in our analysis.
To find out whether the rising incidence is really a
global phenomenon, we carried out an analysis of in-
cidence trends among 37 populations worldwide for
which the data had been collected for 8 years or
more. The incidence of Type I diabetes is globally in-
creasing by 3.0% per year (or by 2.5%, when the in-
cidences were weighted by the number of diabetic
children included in the individual studies). Confi-
dence intervals for these estimates were fairly narrow
indicating that these estimates are reliable. Accord-
ing to this estimate, the incidence of Type I diabetes
will be 40% higher in 2010 than in 1998. This is a real-
istic, although a rather frightening, scenario.
P. Onkamo et al.: Worldwide increase in incidence of Type I diabetes 1399
Fig. 1. Trends in incidence of Type I diabetes in European pop-
ulations. AUT: Austria; BGR: Bulgaria; DEU: East-Germany;
EST: Estonia; FIN: Finland; FRA: France; HUN: Hungary;
ICE: Iceland; ITA: Italy (Turin); LVA: Latvia; LTU: Lithuania;
MAT: Malta; NOR: Norway; POLk: Poland (Krakow); POLr:
Poland (Rzeszów); POLw: Poland (Wielkopolska); SVK: Slo-
vakia; SWE: Sweden; UNKl (UK, Leicestershire); UNKo (Ox-
ford); UNKp (Plymouth); UNKs (Scotland); UNKy (York-
shire). The model fitted to the incidence data was a multiplica-
tive regression model with logarithm of the age standardized
incidence as dependent variable, thus the scale of the incidence
is logarithmic when straight lines were used in drawing the re-
gression lines
Fig. 2. Trends in the incidence of Type I diabetes mellitus in
Non-European populations. DZD: Algeria (Oran); AUSw:
Australia (West); CANm: Canada (Montreal); CANp: Canada
(Prince Edward Island); CHN: China (Shanghai); ISR: Israel
(Yemenite Jews); JPN: Japan (Hokkaido); LIY: Libya; NEZa:
New Zealand (Auckland); NEZc: New Zealand (Canterbury);
PER: Peru (Lima); USAa: USA (Allegheny County); USAc:
USA (Colorado); USAh: USA (Hawaii)
The global variation in the incidence of Type I dia-
betes is prominent [10, 13, 14]. It reflects the distribu-
tion of ethnic diversity showing the importance of the
differential genetic susceptibility among populations.
The incidence is higher among Caucasoid popula-
tions than among Mongoloids and Blacks. Within
ethnic groups, however, there are geographical differ-
ences in incidence depending on the admixture be-
tween racial groups and possible environmental ex-
posures [13]. Although most of the populations in-
cluded in this analysis were Caucasoid, statistically
significant increases in incidence were also found
among the Asian populations in China and Japan,
Mestizos in Peru and also among the Polynesians in
Hawaii. In this literature review it was not possible
to account for ethnic differences within populations
because the authors of the original papers had usually
not given detailed information on incidence in differ-
ent ethnic groups. Where the ethnic groups had been
analyzed separately, however, an increase in the inci-
dence had been observed in all groups, but the rate
of increase could vary from one ethnic group to an-
other. Overall, the increase in the incidence of Type
I diabetes does not seem to be restricted to any par-
ticular ethnic group.
In most countries with a low incidence the stan-
dardized incidence data have been collected during a
relatively short period, which may in some cases ex-
plain the large relative change in incidence. The inci-
dence of Type I diabetes possibly has been underesti-
mated in earlier studies because of incomplete case-
ascertainment and death from undiagnosed diabetes.
Among those populations where the study period
was 18 years or more the increase in incidence was
usually low (from 1.5 to 3.2% per year except Japan,
Hokkaido, 5.9%). Therefore, results from several in-
dividual populations showing large increases should
be interpreted cautiously when the number of cases
P. Onkamo et al.: Worldwide increase in incidence of Type I diabetes1400
Fig. 3. Association between the increase in incidence and the
level of incidence of Type I diabetes. The correlation coeffi-
cient between the log-transformed level of incidence predicted
for 1983 and the increase in incidence was ±0.56 (p = 0.0004)
for all populations. The increase in incidence was estimated
from multiplicative regression models. For each population
the level of incidence was calculated as the model-predicted
(fitted) incidence in 1983 according to the multiplicative re-
gression model. Leicestershire, UK, has been excluded. DZD:
Algeria (Oran); AUSw: Australia (West); AUT: Austria;
BGR: Bulgaria; CANm: Canada (Montreal); CANp: Canada
(Prince Edward Island); CHN: China (Shanghai); DEU: East-
Germany; EST: Estonia; FIN: Finland; FRA: France; HUN:
Hungary; ICE: Iceland; ISR: Israel (Yemenite Jews); ITA: Ita-
ly (Turin); JPN: Japan (Hokkaido); LVA: Latvia; LIY: Libya;
LTU: Lithuania; MAT: Malta; NEZa: New Zealand (Auck-
land); NEZc: New Zealand (Canterbury); NOR: Norway;
PER: Peru (Lima); POLk: Poland (Krakow); POLr: Poland
(Rzeszów); POLw: Poland (Wielkopolska); SVK: Slovakia;
SWE: Sweden; UNKo: UK (Oxford); UNKp: UK (Plymouth);
UNKs: UK (Scotland); UNKy: UK (Yorkshire); USAa: USA
(Allegheny County); USAh: USA (Hawaii); USAc: USA
(Colorado)
Table 3. Predicted incidence (100000 per year) of Type I dia-
betes in children aged 14 years or less by the year 2010 accord-
ing to both the multiplicative and additive regression model.
The populations are arranged in descending order according
to the predicted incidence in 2010 based on the additive model.
Only the populations with a study period covering 14 years or
more were included in predictions with the multiplicative mod-
el
Country: area Predicted incidence per
100000/year in 2010
Additive
regression
model
Multiplicative
regression
model
Finland 50.2 57.9
Norway 41.8
Canada: Prince Edward Island 39.2 51.4
Australia: West 36.7
UK: Scotland 34.9 40.0
UK: Oxford 33.0
Sweden 32.2 33.7
New Zealand: Auckland 27.7
New Zealand: Canterbury 25.0
UK: Plymouth 23.4 27.0
USA: Allegheny County 22.5 24.5
USA: Hawaii 21.4
UK: Yorkshire 21.0 23.0
Hungary 20.1
Libya 19.0
Malta 16.8 15.8
Iceland 16.1 16.0
Slovakia 16.0
Algeria: Oran 15.2
Canada: Montreal 14.8 15.0
France 13.7
Poland: Krakow 13.0
Italy: Turin 12.8
Austria 12.8 14.6
East Germany 12.0 15.4
USA: Colorado 12.0
Poland: Wielkopolska 11.8 20.2
Estonia 10.9 11.0
Bulgaria: East 10.3 10.7
Israel: Yemenite Jews 10.1 12.9
Latvia 8.9
Lithuania 7.5
Poland: Rzeszów 6.7
Japan: Hokkaido 4.1 7.9
China: Shanghai 1.7 3.3
Peru: Lima 1.3
is small and the study period short. The analysis of
the pooled data was repeated excluding populations
for which the case ascertainment level was not report-
ed (Algeria, Israel, Libya, Malta and Oxford, UK).
The results were essentially the same as in the analy-
sis using all data: 2.95% per year vs 2.96% per year,
respectively, and for the weighted regression, 2.40%
vs 2.49%, respectively.
There is presently no way to know whether the ob-
served trend in incidence might reflect a change in
the age at onset of diabetes instead of a true rise in
prevalence. The increase in incidence in 0±14-year-
olds might just be a transition of the age at onset
from the age group 15 years or older. Data on inci-
dence trends in older age groups exist from just a
few populations; thus, reliable information a possible
decrease in incidence in young adults is not available.
Our main result is that the incidence is globally in-
creasing in the age group of 0±14-year-olds.
The genetic factors have been shown to be impor-
tant in the susceptibility to Type I diabetes [5, 6]. Al-
though it is possible that the part of the population
genetically predisposed for Type I diabetes is increas-
ing, this increase may have been modest during the
last decades and not alone a sufficient cause for the
observed increase in incidence. The changes in the
genetic code of the human populations are usually
slow. In this analysis even the longest study period
covered only 30 years, which is approximately the
time span of one generation. It is very unlikely that a
three to tenfold increase in the proportion of subjects
with genetic susceptibility to Type I diabetes has tak-
en place in any population during such a short time.
Instead, the penetrance of the susceptibility genes
might be changing. The penetrance is likely to be de-
termined by an interaction between several suscepti-
bility genes and unknown environmental factors [4,
24].
During the recent years much attention has been
paid to the identification and possible control of envi-
ronmental factors which possibly initiate or trigger
the process leading to Type I diabetes. Although
some studies suggest associations between environ-
mental factors such as diet and viral infections with
the risk of Type I diabetes [25±35], their causative
role has not been shown. It is also difficult to show
that any of these environmental factors has changed
in such a way that a continuous global increase in the
incidenceof TypeI diabeteswould beeasily explained.
The incidence of Type I diabetes is increasing
worldwide. Thus far no population has been identi-
fied in which the incidence has significantly de-
creased. The population-based WHO DIAMOND
Project and the EURODIAB study started at the be-
ginning of the 1990s but have not yet reported results
from the long-term progress in the incidence of Type
I diabetes. It seems obvious that in both of these stud-
ies the 10-year monitoring period planned thus far is
too short to produce reliable trend estimations and
predictions for the change of the incidence of Type I
diabetes, especially in countries where incidence is
low. There is a need to continue with the communi-
ty-based registries on Type I diabetes worldwide. Ef-
forts are also needed to identify effective primary
prevention measures for Type I diabetes to stop the
global increase in the incidence of this disease.
Acknowledgements. We thank the Diabetes Epidemiology Re-
search International (DERI) Group for the data for several
countries used in this study.
References
1. Cudworth AG, Wolf E, Gorsuch AN, Festenstein H (1979)
A new look at HLA genetics with particular reference to
type-1 diabetes. Lancet 2: 389±391
2. Tuomilehto-Wolf E, Tuomilehto J (1991) HLA antigens in
insulin-dependent diabetes mellitus. Ann Med 23: 481±
488
3. Dahlquist G, Mustonen L (1994) Childhood onset diabetes
± time trends and climatological factors. Int J Epidemiol 23:
1234±1241
4. Todd JA (1997) Genetics of type 1 diabetes. Pathol Biol
(Paris) 45: 219±227
5. Kaprio J, Tuomilehto J, Koskenvuo M et al. (1992) Concor-
dance for type 1 (insulin-dependent) and type 2 (non-insu-
lin-dependent) diabetes mellitus in a population-based co-
hort of twins in Finland. Diabetologia 35: 1060±1067
6. Kyvik KO, Green A, Beck-Nielsen H (1995) Concordance
rates of insulin dependent diabetes mellitus: a population
based study of young Danish twins. BMJ 311: 913±917
7. Blom L, Nyström L, Dahlquist GG (1991) The Swedish
childhood diabetes study: vaccinations and infections as
risk determinants for diabetes in childhood. Diabetologia
34: 176±181
8. Virtanen SM, Räsänen L, Aro A et al. (1992) Feeding in in-
fancy and the risk of type 1 diabetes mellitus in Finnish
children. Diabet Med 9: 815±819
9. LaPorte RE, Tajima N, kerblom HK et al. (1985) Geo-
graphic differences in the risk of insulin-dependent diabe-
tes mellitus: the importance of registries. Diabetes Care
8[Suppl 1]:S101±S107
10. Rewers M, LaPorte RE, King H, Tuomilehto J (1988)
Trends in the prevalence and incidence of diabetes: insu-
lin-dependent diabetes mellitus in childhood. World
Health Stat Q 41: 179±189
11. WHO DIAMOND Project Group on Epidemics (1992)
Childhood diabetes, epidemics, and epidemiology: an ap-
proach for controlling diabetes. Am J Epidemiol 135:
803±816
12. Green A, Gale EA, Patterson CC for the EURODIAB
ACE study (1992) Incidence of childhood-onset insulin-de-
pendent diabetes mellitus: the EURODIAB ACE study.
Lancet 339: 905±909
13. Karvonen M, Tuomilehto J, Libman I, LaPorte R for the
World Health Organization DIAMOND Project Group
(1993) A review of the recent epidemiological data on the
worldwide incidence of type 1 (insulin-dependent) diabetes
mellitus. Diabetologia 36: 883±892
14. Karvonen M, Pitkäniemi M, Pitkäniemi J, Kohtamäki K,
Tajima N, Tuomilehto J for the World Health Organization
P. Onkamo et al.: Worldwide increase in incidence of Type I diabetes 1401
DIAMOND Project Group (1997) Sex difference in the in-
cidence of insulin-dependent diabetes mellitus: an analysis
of the recent epidemiological data. Diabetes Metab Rev
13: 275±291
15. Diabetes Epidemiology Research International Group
(1990) Secular trends in incidence of childhood IDDM in
10 countries. Diabetes 39: 858±864
16. Diabetes Epidemiology Research International Group
(1988) Geographic patterns of childhood insulin-depen-
dent diabetes mellitus. Diabetes 37: 1113±1119
17. Schoenle EJ, Molinari L, Bagot M, Semadeni S, Wiesen-
danger M (1994) Epidemiology of IDDM in Switzerland.
Increasing incidence rate and rural-urban differences in
Swiss men born 1948±1972. Diabetes Care 17: 955±960
18. Ruwaard D, Gijsen R, Bartelds IM, Hirasing RA, Verkleij
H, Kromhaut D (1996) Is the incidence of diabetes increas-
ing in all age-groups in The Netherlands? Results of the
second study in the Dutch Sentinel Practice Network. Dia-
betes Care 19: 214±218
19. Bingley PJ, Gale EA (1989) Rising incidence of IDDM in
Europe. Diabetes Care 12: 289±295
20. Green A, Andersen PK, Svendsen AJ, Mortensen K (1992)
Increasing incidence of early onset type 1 (insulin-depen-
dent) diabetes mellitus: a study of Danish male birth co-
horts. Diabetologia 35: 178±182
21. Jaksic J, Matic I, Stojnic E, Juros A, Pelajic A (1996) Inci-
dence of insulin dependent diabetes mellitus in children
aged 0±19 in the Sibenic area. Diabetol Croat 29±33
22. Choubnikova J, Shubnikof E, Kalashnikova L (1996) The
increase of type 1 diabetes incidence among children in
Novosibirsk city. 32nd Annual Meeting of the European
Association for the Study of Diabetes, Vienna (Abstract)
23. Shaltout AA, Qeabazard MA, Abdella et al. (1995) High
incidence of childhood-onset IDDM in Kuwait. Diabetes
Care 18: 923±927
24. Cordell HJ, Todd JA (1995) Multifactorial inheritance in
type 1 diabetes. Trends Genet 11: 499±504
25. Borch-Johnsen K, Joner G, Mandrup-Poulsen T et al.
(1984) Relation between breast-feeding and incidence
rates of insulin-dependent diabetes mellitus. A hypothesis.
Lancet 2: 1083±1086
26. Kostraba JN, Cruickshanks KJ, Lawler-Heavner J et al.
(1993) Early exposure to cow's milk and solid foods in in-
fancy, genetic predisposition, and risk of IDDM. Diabetes
42: 288±295
27. Elliott RB, Martin JM (1984) Dietary protein: a trigger of
insulin-dependent diabetes in the BB rat? Diabetologia
26: 297±299
28. Helgason T, Jonasson MR (1981) Evidence for a food addi-
tive as a cause of ketosis-prone diabetes. Lancet 2: 716±720
29. Dahlquist GG, Blom LG, Persson LA, Sandström AI, Wall
SG (1990) Dietary factors and the risk of developing insulin
dependent diabetes in childhood. BMJ 300: 1302±1306
30. Virtanen SM, Jaakkola L, Räsänen L et al. (1994) Nitrate
and nitrite intake and the risk for type 1 diabetes in Finnish
children. Diabet Med 11: 656±662
31. Rayfield EJ, Seto Y (1978) Viruses and the pathogenesis of
diabetes mellitus. Diabetes 27: 1126±1140
32. Gamble DR (1976) A possible virus etiology for juvenile
diabetes. In: Creutzfeld W, Köbberlin J, Neel V (eds) The
genetics of diabetes mellitus. Springer, Berlin Heidelberg
New York, pp 95±105
33. Hyöty H, Leinikki P, Reunanen A et al. (1988) Mumps in-
fections in the etiology of type 1 (insulin-dependent) diabe-
tes. Diabetes Res 9: 111±116
34. Hyöty H, Hiltunen M, Knip M et al. (1995) A prospective
study of the role of coxsackie B and other enterovirus in-
fections in the pathogenesis of IDDM. Diabetes 44:
652±657
35. Hiltunen M, Hyöty H, Karjalainen J et al. (1995) Serologi-
cal evaluation of the role of cytomegalovirus in the patho-
genesis of IDDM: a prospective study. Diabetologia 38:
705±710
36. Bessaoud K, Boudraa G, Deschamps I, Hors J, Benbouab-
dallah M, Touhami M (1990) Epidemiologie du diabte in-
sulinodØpendant juvØline en AlgØrie (Wilaya D'Oran).
Rev Epidemiol Sante Publique 38: 91±99
37. Kelly HA, Russell MT, Jones TW, Byrne GC (1994) Dra-
matic increase in incidence of insulin dependent diabetes
mellitus in Western Australia. Med J Aust 161: 426±429
38. Schober E, Schneider E, Waldhör T, Tuomilehto J and
Austrian Diabetes Incidence Study Group (1995) Increas-
ing incidence of IDDM in Austrian children. A nationwide
study 1979±1993. Diabetes Care 18: 1280±1283
39. Tzaneva V, Iotova V, Bruining GJ (1998) Increase in
IDDM incidence in Bulgarian children (1974±1995). J Pe-
diatr Endocrinol Metab 11: 725±732
40. Siemiatycki J, Colle E, Campbell S, Dewar R, Aubert D,
Belmonte MM (1988) Incidence of IDDM in Montreal by
ethnic group and by social class and comparisons with eth-
nic groups living elsewhere. Diabetes 37: 1096±1102
41. Tan MH, Wornell MC, Beck AW (1981) Epidemiology of
diabetes mellitus in Prince Edward Island. Diabetes Care
4: 519±524
42. Väänänen S, Kohtamäki K, Karvonen M, Tuomilehto J for
the Diamond Research Group (1997) Incidence of IDDM
Worldwide Preliminary Analysis of the WHO DIAMOND
Incidence Data. International Diabetes Epidemiology
Group Symposium. Savonlinna, Finland (Abstract)
43. Fu H, Shen SX, Chen ZW et al. (1994) Shanghai, China,
has the lowest confirmed incidence of childhood diabetes
in the world. Diabetes Care 17: 1206±1208
44. Shen SX, Wang HB, Chen ZW et al. (1996) The incidence
of insulin-dependent diabetes mellitus in urban districts of
Shanghai (1989±1993). J Pediatr Endocrinol Metab 9:
469±473
45. Michaelis D, Jutzi E, Vogt L (1993) Epidemiology of insu-
lin-treated diabetes mellitus in the East-German popula-
tion: differences in long-term trends between incidence
and prevalence rates. Diabetes Metab 19: 110±115
46. Podar T, LaPorte RE (1993) Incidence of childhood diabe-
tes did not increase in Estonia during 1980±89. Diabetes
Metab 19: 361±363
47. Tuomilehto J, Virtala E, Karvonen M et al. (1995) Increase
in incidence of insulin-dependent diabetes mellitus among
children in Finland. Int J Epidemiol 24: 984±992
48. LØvy-Marchal C, and Groupe du Registre d'incidence
du Diabte Insulino-DØpendant de l'Enfant (1998) Évolu-
tion de l'incidence du diabte insulino-dØpendant de l'en-
fant en France. Rev Epidemiol Sante Publique 46: 157±163
49. SoltØsz G, Madµcsy L, BØkefi D, Dankó I, the Hungarian
Childhood Diabetes Epidemiology Group (1990) Rising
incidence of Type 1 diabetes in Hungarian children
(1978±1987). Diabet Med 7: 111±114
50. Helgason T, Danielsen R, Thorsson AV (1992) Incidence
and prevalence of type 1 (insulin-dependent) diabetes mel-
litus in Icelandic children 1970±1989. Diabetologia 35:
880±883
51. Shamis I, Gordon O, Albag Y, Goldsand G, Laron Z (1997)
Ethnic differences in the incidence of childhood IDDM in
Israel (1965±1993) Marked increase since 1985, especially
in Yemenite Jews. Diabetes Care 20: 504±508
52. Bruno G, Merletti F, De Salvia A, Lezo A, Arcari R, Pa-
gano G, Piedmont Study Group for Diabetes Epidemiolo-
P. Onkamo et al.: Worldwide increase in incidence of Type I diabetes1402
gy (1997) Comparison of incidence of insulin-dependent di-
abetes mellitus in children and young adults in the province
of Turin, Italy, 1984±91. Diabet Med 14: 964±969
53. Matsuura N, Fukuda K, Okuno A et al. (1998) Descriptive
epidemiology of IDDM in Hokkaido, Japan: the Child-
hood IDDM Hokkaido Registry. Diabetes Care 21:
1632±1636
54. Padaiga Z, Tuomilehto J, Karvonen M et al. (1997) Inci-
dence trends in childhood onset IDDM in four countries
around the Baltic sea during 1983±1992. Diabetologia 40:
187±192
55. Brigis G, Robeza I (1997) Childhood onset IDDM inci-
dence rates in Latvia. Outbreak in 1995, comparisons by
ethnicity International Diabetes Epidemiology Group
Symposium. Savonlinna, Finland (Abstract)
56. Kadiki OA, Moawad SE (1993) Incidence and prevalence
of type 1 diabetes in children and adolescents in Benghazi,
Libya. Diabet Med 10: 866±869
57. Schranz AG, Prikatsky V (1989) Type 1 diabetes in the
Maltese Islands. Diabet Med 6: 228±231
58. Elliott RB, Pilcher CC (1985) Childhood diabetes in the
Auckland area. N Z Med J 98: 922±923
59. Scott RS, Brown LJ, Darlow BA, Frobes LV, Moore MP
(1992) Temporal variation in incidence of IDDM in Can-
terbury, New Zealand. Diabetes Care 15: 895±899
60. Joner G, Sùvik O (1989) Increasing incidence of diabetes
mellitus in Norwegian children 0±14 years of age
1973±1982. Diabetologia 32: 79±83
61. SeclØn S, Rojas MI, Nuæez H, Valdivia H, Millones B, Dia-
betes Epidemiology Research Peruvian Group (DERPG)
(1997) Type 1 (insulin-dependent) diabetes in Mestizo chil-
dren of Lima, Peru. Report on a ten years (1985±94) inci-
dence. International Diabetes Epidemiology Group Sym-
posium. Savonlinna, Finland. (Abstract)
62. Szybìnski Z, Czyzyk A, Wasik R, Dziatkowiak H, Cie-
chanowska M, Symonides-Lawecka A, Szurkowska DP
(1997) Epidemiology of IDDM in Poland ± a 10 year per-
spective International Diabetes Epidemiology Group Sym-
posium. Savonlinna, Finland (Abstract)
63. Grzywa MA, Sobel AK (1995) Incidence of IDDM in the
province of Rzeszów, Poland, 0- to 29-year-old age-group,
1980±1992. Diabetes Care 18: 542±544
64. Rewers M, Stone RA, LaPorte RE et al. (1989) Poisson re-
gression modeling of temporal variation in incidence of
childhood insulin-dependent diabetes mellitus in Alleghe-
ny County, Pennsylvania, Wielkopolska, Poland, 1970±
1985. Am J Epidemiol 129: 569±581
65. Michalkovµ DM, Cernay J, Dankovà A, Rusnµk M,
Fandµkova K, Slovak Childhood Diabetes Epidemiology
Study Group (1995) Incidence and prevalence of childhood
diabetes in Slovakia (1985±1992). Diabetes Care 18:
315±320
66. Michalkova D, Barak L, Dankova A, Pastor K, Mikulecky
M, Cernay J, Slovak Pediatric Epidemiology Group
(1997) The incidence and prevalence of type 1 diabetes
mellitus in Slovak children from 1985 to 1995. International
Diabetes Epidemiology Group Symposium. Savonlinna,
Finland (Abstract)
67. Gardner SG, Bingley PJ, Sawtell PA, Weeks S, Gale EA,
the Bart's-Oxford Study Group (1997) Rising incidence of
insulin dependent diabetes in children aged under 5 years
in the Oxford region: time trend analysis. BMJ 315:
713±717
68. Zhao HX, Stenhouse E, Soper C et al. (1999) Incidence of
childhood onset Type I diabetes mellitus in Devon and
Cornwall, England, 1975±1996. Diabetic Med 16: (in press)
69. Staines A, Bodansky HJ, Lilley HE, Stephenson C, McNal-
ly RJ, Cartwright RA (1993) The epidemiology of diabetes
mellitus in the United Kingdom: The Yorkshire Regional
Childhood Diabetes Register. Diabetologia 36: 1282±1287
70. McKinney PA, Law G, Staines AS, Bodansky HJ (1995) In-
cidence of diabetes in children. BMJ 310: 1672
71. LaPorte RE, Fishbein HA, Drash AL et al. (1981) The
Pittsburgh insulin-dependent diabetes mellitus (IDDM)
registry: the incidence of insulin-dependent diabetes melli-
tus in Allegheny County, Pennsylvania (1965±1976). Dia-
betes 30: 279±284
72. Kostraba JN, Gay EC, Cai Yet al. (1992) Incidence of insu-
lin-dependent diabetes mellitus in Colorado. Epidemiolo-
gy 3: 232±238
73. Patrick SL, Kadohiro JK, Waxman SH et al. (1997) IDDM
Incidence in a Multiracial Population. The Hawaii IDDM
Registry, 1980±1990. Diabetes Care 20: 983±987
P. Onkamo et al.: Worldwide increase in incidence of Type I diabetes 1403