Vol.:(0123456789)Diabetologia
https://doi.org/10.1007/s00125-026-06675-9
ARTICLE
Geospatial clustering of type 1 diabetes in Sweden: a cohort study based on all residential locations from birth to diagnosis
Samy Sebraoui 1 · Oskar Englund2 · Fredrik Nyberg3 · Annelie Carlsson4 · Olle Korsgren5 · Gun Forsander6 ·
Katarina Eeg‑Olofsson1 · Björn Eliasson7 · Hanne K. Carlsen8 · Karin Åkesson8,9 · Soffia Gudbjörnsdottir
Vol.:(0123456789)Diabetologia
https://doi.org/10.1007/s00125-026-06675-9
ABSTRACT
Aims/hypothesis
Type 1 diabetes develops gradually, and previous exposures may influence incidence. We aimed to assess
the geographical variation in type 1 diabetes incidence in Sweden by considering all residential locations from birth to
diagnosis in individuals aged 0–30 years, diagnosed between 2005 and 2022. Significant high- and low-risk clusters were
identified for different life stage exposure windows.
Methods
In 21,774 individuals with type 1 diabetes, all residential geographical locations from birth to diagnosis were
geocoded. Geostatistical analysis of the incidence of type 1 diabetes was conducted at the municipality level using the most
common residential location during four life stage-specific exposure windows (at diagnosis, the first 5 years after birth, 5
years prior to diagnosis, and from birth to diagnosis).
Spatial scan statistics were used to identify statistically significant
high- and low-risk clusters for each window. Land use and land cover within these clusters were also characterised.
Results
Significant geographical variation in the incidence of type 1 diabetes was observed.
The incidence was consistently higher in rural, low-population-density areas, particularly in central Sweden, and lower in major urban areas.
The largest number of spatial clusters of both high risk (RR 1.29–16.0) and low risk (RR 0.32–0.73) was identified when using the most
common residential location during the first 5 years after birth. High-risk clusters for this exposure window were charac-
terised by forested and agricultural land, while low-risk clusters were characterised by urban land and open land other than
agricultural land.
Conclusions/interpretation
Our findings suggest that the development of type 1 diabetes in Sweden varies geographically
and is associated with specific features of the local surroundings in early childhood.
This is important knowledge as a basis
for identifying possible environmental risk factors and the relationship with risk of type 1 diabetes in future studies.
From the article
Discussion
In this nationwide study of individuals diagnosed with type
1 diabetes in Sweden between 2005 and 2022 at the age
of 0–30 years, we identified a notable geographical varia-
tion in incidence. Using high-resolution geospatial analyses,
we observed that the incidence of type 1 diabetes was con-
sistently higher in rural and low-population-density areas
particularly in the central part of the country and lower in
major urban areas. These patterns were consistent using
various analytical approaches and between four different life
stage-specific exposure windows. The most striking find-
ings emerged when evaluating the most dominant residential
location in the first 5 years after birth, which revealed the
highest number of significant high-risk and low-risk spatial
clusters.
Nearly one in four individuals in this study relocated
between municipalities between birth and diagnosis,
underscoring the potential for exposure misinterpretation
in studies relying solely on residential location at the time
of diagnosis or even later [10, 13, 14, 17]. By incorporat-
ing longitudinal residential data and analysing different life
stage exposure windows, our approach captures critical
periods of the development of the disease, including early-
life exposure, which may contribute differently to type 1
diabetes pathogenesis. Patients in the high-risk clusters
were less likely to relocate to a different municipality com-
pared with those in the low-risk clusters, both from birth to
diagnosis (26.4% vs 38.7%) and during the first 5 years after
birth (25.1% vs 28.4%). This may have contributed to the
stronger association between place of birth and subsequent
incidence.
Across the 290 municipalities the incidence rate varied
more than sixfold from 13.8 to 91.1 per 100,000 person-
years. The highest incidence was found predominantly in
sparsely populated municipalities located in central Swe-
den, without a clear north-south gradient, in contrast to the
findings of previous studies [31]. Conversely, the lowes
risk was observed in urban areas around the largest cities.
These results are in line with previous data from the Nor-
dic countries of children 0–14 years old during 2006–2011
[32], as well as with an older study from Finland [33]. Our
study extends these findings across a wider age span and
over a longer period. Most importantly, our findings suggest
that the risk associated with geographical location is more
pronounced when considering early-life residential location
rather than location at diagnosis.
Both the risk of type 1 diabetes and the residential distri-
bution of individuals with non-Swedish ethnicity vary across
the country. A previous study showed that children in fami-
lies that have immigrated to Sweden have lower incidence
compared with Sweden-born children [34]. These factors
cannot explain our findings since separate analysis for indi-
viduals born in Sweden to two Sweden-born parents showed
the same general geographical patterns for all life stage
exposure windows when compared with the analysis with
all cases included.
However, municipalities that previously
demonstrated low incidence when all cases were included,
particularly in the southern regions of the country, showed
higher incidence when the non-native group was excluded.
This phenomenon could be explained by immigration.
The study of geographical variation based on munici-
palities is limited as some municipalities are very sparsely
populated and only a few cases might make a large differ-
ence, and the large number of municipalities implies a risk
of significance by chance. In addition, artificial boundaries
that do not account for geographical or environmental fac-
tors may result in overlooking accumulations of geographi-
cal influences that are spread across multiple municipalities.
To overcome these limitations, we also identified clusters
independently of any artificial boundaries. This approach
provided a more detailed analysis compared with methods
based on municipal boundaries and should result in more
accurate geographical patterns.
High-risk and low-risk clusters were analysed for the four
life stage exposure windows. Again, this approach is essen-
tial given that the disease takes years to develop [11] and
that one fourth of all individuals relocated between birth
and diagnosis. However, no earlier studies have assessed
geographical exposure at different life stages leading up to
the development of the disease.
The cluster analysis at diagnosis revealed four significant
high-risk clusters with an RR from 1.31 to 1.80 located in
the mid country, and five low-risk clusters in the largest cit-
ies, which largely mirrored the results from the municipal-
ity-based analysis. Country-specific incidence rate studies
have shown rural excess in some countries of type 1 diabetes
[10, 12, 15, 16, 35] but not all [13, 36–38]. These studies
vary in terms of methodology, and some studies only have
a few cases and small geographical units. The characteris-
tics of rural/urban environments also need to be considered.
Several studies have found a higher incidence of type 1 dia-
betes in areas with lower levels of deprivation [35, 39–42].
A study from Sweden has shown that low maternal educa-
tion increases the risk of the disease [43]. In the current
study, the observed differences in incidence between rural
and urban areas could be attributed to spatial patterns in the
population composition of individual-level socioeconomic
status, which may be related to differences in lifestyle, expo-
sure to infections or other exposures.
The LULC analyses revealed a distinct urban–rural con-
trast: high-risk clusters were largely in rural areas, with a
high proportion of forested and agricultural land; low-risk
clusters were dominated by urban land and open land other
than agriculture. Given that the highest number of significant
high-risk and low-risk spatial clusters were found for the
first 5 years after birth, this suggests that the risk for type
1 diabetes could be linked to specific features of the local
surroundings in early childhood, independently of the age
of diagnoses. These could be associated with environmental
risk factors specific to rural living or protective factors in
urban living during early childhood. Our results are consist-
ent with growing evidence suggesting that early-life events
including parental psychosocial stress, pregnancy-related
factors, infant growth patterns, infections, and environmen-
tal exposures may influence the later development of type
1 diabetes [8, 9, 44, 45]. This is important knowledge to be
able to identify possible early environmental risk factors of
type 1 diabetes in future studies. Our finding of lower risk
in urban areas is in line with a recent study from England,
which reported that air pollution, light at night, population
density, and overcrowding were negatively associated with
type 1 diabetes incidence [10].
Furthermore, day care attendance may be associated
with a reduced risk of type 1 diabetes [46]. Our findings
could imply the presence of a protective effect linked to
characteristics of the urban environment and might partly
be explained by the hygiene hypothesis, which posits that
children with low exposure to infectious agents at an early
stage after birth have an increased susceptibility for type 1
diabetes [44, 47–49].
It is unlikely that a difference in geographical distribu-
tion of high-risk HLA haplotypes could explain our finding,
especially given that less than 5% of individuals with these
haplotypes in the general population develop overt type 1
diabetes, strongly suggesting a second critical hit is needed
[50].
The strengths of this study are that it is a large nation-
wide study from a country with a high incidence of type
1 diabetes, covering both children and young adults. We
used precise geographical coordinates of all cases from
birth to the year of diagnosis. High-resolution geospatial
analyses were used to identify geographical incidence pat-
terns during different life stage exposure windows. This is
unique and has not been previously conducted in diabetes
epidemiology. All analyses were compared with those of
the mean population in the same age group, during the
study period, and thus, the individuals with type 1 diabetes
were consistently matched with those in the corresponding
population group. We have also quantitatively character-
ised all high- and low-risk clusters in terms of land use/
land cover.
The weaknesses of the study are that spatial analysis is
sensitive to sample size, which implies that areas with a
small number of type 1 diabetes cases may result in the iden-
tification of large areas as ‘hot spots’ or clusters. Given the
large study population and the long observation time, this
should not affect our conclusions. The coverage of NDR was
somewhat lower during the first 2 years of the study period.
However, NDR collects repeated registrations (including
date or year of diagnosis) at least annually; therefore, almost
all patients with type 1 diabetes diagnosed between 2005 and
2022 are eventually captured. We lacked access to gridded
population data for the period 2005–2014. Consequently,
it was necessary to interpolate population estimates based
on municipal population data from 2015–2022, which were
subsequently transferred to the grid cells. Municipal popu-
lation data over time exhibited minimal variation. Type 1
diabetes is a relatively rare disease, and robust cluster analy-
ses require large patient numbers to ensure statistical cer-
tainty. Subdivisions in time periods or age groups further
reduce numbers and power; thus, such data could be prone
to spurious findings. Furthermore, we did not assess whether
the results differed between male and female individuals or
explore gender-related factors, which limits generalisability
across all sexes and genders.
We deliberately chose to use crude incidence rate to
ensure methodological consistency and comparability across
the difference analyses. Sensitivity analyses did not show
any substantial differences between age- and sex-standard-
ised incidence rates and crude incidence. Thus, the methodi-
cal choice is highly unlikely to affect our conclusions (ESM
Fig. 4).
Conclusion
Type 1 diabetes incidence in Sweden shows
clear geographical variation, with higher rates in rural, low-
population-density areas, particularly in central Sweden,
and lower rates in major urban areas. The strongest spatial
patterns were linked to residence in early childhood, sug-
gesting that local environmental features – such as forested
and agricultural land – may influence disease risk. These
findings underscore the importance of investigating early-life
environmental exposures in future type 1 diabetes research.
Nyhetsinfo
Läs hela artrikeln open source pdf free
https://pubmed.ncbi.nlm.nih.gov/41692841/
AI
2026 Swedish cohort study found that type 1 diabetes (T1D) incidence is higher in rural areas and lower in major cities, with significant spatial clustering linked to residential location, especially during the first 5 years of life
. High-risk clusters (1.29–16.0 RR) were associated with forested/agricultural land, while low-risk clusters (0.32–0.73 RR) were in urban areas. springermedicine.com +2
Key Findings:
Early Life Exposure: The highest number of high- and low-risk clusters were identified based on the first 5 years of life, suggesting environmental factors during early childhood strongly influence T1D development.
Urban-Rural Disparity: Significant high-risk clusters for T1D (30% to 80% higher than average) were found in rural, central Sweden.
Protective Urban Factors: Conversely, major cities like Stockholm, Gothenburg, and Malmö showed significant low-risk clusters.
Study Design: The study (2005–2022) analyzed T1D incidence in individuals aged 0–30, accounting for residential mobility, with 24% of patients having moved between municipalities. springermedicine.com +3
These findings suggest that environmental factors, possibly associated with rural living, may trigger the development of T1D, while urban environments might offer protective, yet unidentified, factors. springermedizin.de +1
www red DiabetologNytt