Backgroud
The incidence of diabetes have increased during recent decades [
1,
2]. Studies have shown that the incidence of type 1 diabetes mellitus (T1DM) increased worldwide over the past 3 decades [
3‐
5]. For example, the annual incidence of diabetes among youths increased from 9.0 cases per 100,000 person-year in 2002–2003 to 12.5 cases per 100,000 persons in 2011–2012 in the USA [
6]. The incidence of T1DM differed significantly among European regions, being highest in central and eastern European countries in the 1990s [
7]. The prevalence of being overweight or obese is also increasing worldwide [
8]. The WHO Global Report on Diabetes indicated that being overweight or obese is the strongest risk factor for type 2 diabetes mellitus (T2DM) and that T2DM and prediabetes are increasingly being observed in children, adolescents, and younger adults [
9]. Thus, the increase of overweight rate and obesity rate also affect the incidence of diabetes in different degrees. Diabetes can lead to complications in many parts of the body and can increase disability rates and the occurrence of other complications, resulting in a heavy economic burden. The highest proportion of health-care spending in the USA was on diabetes, costing an estimated $101.4 billion in 2013 [
10]. The incidence of diabetes varies from region to region and is affected by many factors. The human development level of a country was measured using its human development index (HDI): a summary indicator of health, education, and income. The Human Development Index of a region may impact the diabetes incidence locally.
The Global Burden of Disease Study (GBD) has assessed the burden of diabetes mellitus in 194 countries (Taiwan: province of China) and territories around the world and hence provides a unique opportunity to understand the landscape of diabetes mellitus (See Additional file
1 Table S4 for details regarding 194 countries lists) [
11]. In the present study, we retrieved detailed information on the incidence of diabetes mellitus from the GBD performed in 2017. We further assessed the disease burden of diabetes mellitus by determining temporal trends in the incidence of different types of diabetes mellitus from 1990 to 2017 at the global, regional, and national levels. The list of different types of diabetes mellitus has expanded, and many new and more detailed data sources incorporated. We reported the new findings for the first time at the country level for 1990–2017. The findings of the study can assist in the design of targeted strategies for diabetes mellitus prevention tailored to different countries.
Methods
Study data and definitions
According to sex, region, country, and disease type (T1DM and T2DM), the annual incident cases and age-standardized incidence of diabetes from 1990 to 2017 were derived using the Global Health Data Exchange (GHDx) query tool (
http://gdx.healthdata.org/gbd-results-tool) [
12]. GHDx is the world’s most comprehensive catalog of surveys, censuses, vital statistics, and other health-related data. The GBD 2017 study, which included injuries and risk factors, covered 195 countries and territories between 1990 and 2017. In total, 359 diseases and injuries and healthy life expectancy, 282 causes of death and 84 risk factors were systematically analyzed. Data used include vital registration systems, sample registration systems, household surveys (complete birth histories, summary birth histories, sibling histories), censuses (summary birth histories, household deaths), and Demographic Surveillance Sites. The explanation, validation and assessment of data quality have been reported in previous studies [
13‐
15]. The SDI, which is based on national-level income per capita, average years of education among persons older than 15, and total fertility rate, was used to categorize the countries into five SDI quintiles (high, high-medium, medium, low-medium, and low levels). We also collected the human development index (HDI) data at the national level from the Human Development Reports of United nations development programme (
http://hdr.undp.org/en/indicators/137506).
For the definition and classification of diabetes, we refer to GBD research criteria [
13]. The case definitions and diagnostic criteria for overall diabetes mellitus, type 1 diabetes mellitus, and type 2 diabetes mellitus are presented in the Table
1 below.
Table 1
Overall diabetes mellitus, type 1 diabetes mellitus, and type 2 diabetes mellitus
1. Overall diabetes mellitus | Diabetes mellitus (DM) is defined as fasting plasma glucose (FPG) > 126 mg/dL (7 mmol/L) or being on treatment for diabetes. |
2. Overall diabetes mellitus type 1 | Cases of DM that are on insulin or diagnosed with a biomarker (eg, c-peptide levels) that is not fasting plasma glucose |
3. Overall diabetes mellitus type 2 | Cases of diabetes mellitus (DM) type 2 are not reported as type 1 diabetes mellitus. |
Statistical analysis
Analyses were done separately for sex, region and country using a statistical model described and validated previously [
16]. In addition, all the ages were included in the study. The model had a hierarchical structure in which estimates for each country, region and year were informed by its own data, data from other years in the same country and data in other countries in the same region. The model also accounted for non-linear time trends and age associations.
The incidence and prevalence are expressed as age-standardized based on the GBD reference population [
13,
17,
18] unless otherwise specified. We used the age-standardized incidence rate (ASIR) and the estimated annual percentage change (EAPC) to quantify diabetes mellitus incidence trends [
19]. The age-standardized prevalence rate (ASPR) were reported prevalence. ASR(age-standardized rate, include ASIR and ASPR) data can be obtained from the GHDx, with detailed calculation methods available in the literature [
20]. Standardization was necessary for multiple groups of people with different age structures or for the same population in which the age distribution changes over time. The ASR (per 100,000 population) was calculated by summing up the products of the age-specific rates (
ai, where
i denotes the
ith age group) and the number of persons (or weight) (
wi) in the same age group
i of the chosen reference standard population, then dividing by the sum of the standard population weights:
$$ ASR=\frac{\sum_{i=1}^A{a}_i{w}_i}{\sum \limits_{i=1}^A{w}_i}\times \mathrm{100,000} $$
ASIR analysis can be used to better understand the disease models in the population and evaluate the effectiveness of current prevention strategies, and then develop more-targeted strategies where necessary. More importantly, the ASIR trends can serve as a good surrogate for shifting patterns of disease within a population, as well as provide clues about the changing risk factors. Consequently, the effectiveness of current prevention strategies can be assessed, and more-targeted ones can be established (if they are needed) based on the ASIR analyses [
21].
The EAPC was a summary and widely used measure of the ASIR trend over a specified interval and determined by fitting a regression line to the natural logarithm of the ASIR:
y =
α +
βx +
ε, where
y = ln(ASIR) and
x = calendar year. The EAPC was calculated as 100 × (exp(
β) – 1), and its 95% confidence interval (CI) can also be obtained from a linear regression model [
22]. The calendar year was used as a continuous forecast variable. In the present study, the ASIR was deemed to be in an increasing trend if the EAPC and the lower boundary of its 95% CI were both > 0%. In contrast, the ASIR was in a decreasing trend if the EAPC estimation and the upper boundary of its 95% CI were both < 0%; otherwise, the ASIR was deemed to be uncertain over time. Additionally, in order to identify the factors influencing EAPCs, we evaluated the association between EAPC and the HDI in 2017 at the national level.
All statistical analyses were performed using the R program (version 3.5.1). A probability value of p < 0.05 was considered statistically significant.
Discussion
Our study presents the most recent trends and patterns of the worldwide incidence of diabetes mellitus associated with sex, region, country, and types based on data obtained in GBD 2017. In general, both the number of incident cases and the incidence of diabetes mellitus increased from 1990 to 2017, and these trends were dominated by an increase in T2DM, with a smaller contribution from T1DM. The temporal trends in the diabetes incidence varied considerably between different regions and countries, and the heterogeneous pattern in risk-factor exposures resulted in a markedly diverse diabetes mellitus incidence across the world, which indicates the complexity of preventing diabetes mellitus [
23]. This increasing pattern of diabetes mellitus was heterogeneous across regions and countries. This may be related to considerable changes in the population pyramid (age distribution) in some of these countries.
Our finding of an increasing temporal trend of incident diabetes cases over the past two decades is consistent with existing studies. The recent landmark study that performed a pooled analysis across 751 studies involving 4.4 million adults from 200 countries indicated that between 1980 and 2014 the number of adults with diabetes worldwide increased fourfold, from 108 million to 422 million [
24]. The reasons for this upward trend are multiple. Known common factors, such as obesity, physical inactivity, poor dietary habits, hypertension, and dyslipidemia, have been widely reported [
25‐
27].
In addition, other more recently discovered factors, such as intrauterine development [
28], fetal undernutrition, and low birth weight may also contribute to the increasing trend. For example, intrauterine growth restriction may lead to high glucose levels in infants [
29]; diabeteic risk in adulthood is influenced not only by genetic predisposition, but also by environmental factors during early life, such as fetal undernutrition [
30]; low birth weight is also found to be a contributor to early T2DM onset in adulthood [
31]. Given the diversified risk factors, public health prevention and control of diabetes should be customed to the individuals’ needs of the targeted population.
Our finding demonstrates that T2DM individuals account for the vast majority of people living with diabetes worldwide. In developing countries, the increasing T2DM level is largely associated with improvements in social development and living standards, resulting in excessive energy intake and reduced exercise in recent years [
32]. Using the most populous developing country, China, as an example, the prevalence of T2DM in the Chinese population was 7.7% in 2000–2004, 9.3% in 2005–2009 and 10.1% in 2010–2014 from a meta-analysis of research [
33]. Estimated 26.1 million (or 5.5%) adults aged 35 to 74 years in 2000–2001 and 118.5 million(or 10.9%) Chinese adults had diabetes in 2013 [
34,
35]. Consistently, China’ Per Capita GDP has increased from 7078 yuan(RMB) to 46,629 yuan(RMB) during 2000–2014 from National Bureau of Statistics of China. The significant increase in the incidence of diabetes in China is closely related to its economic development. Besides, the age of T2DM onset also become younger due to changing in dietary habits and sedentary lifestyle related to economic development.
Our review of the worldwide epidemiology of T1DM revealed that its incidence varied markedly across countries and regions worldwide. The ASIR of T1DM differed with latitude and the farther away from the equator, the higher the ASIR in 2017. A particularly interesting finding in our study was that while the ASIR of T1DM decreased over the last two decades overall in Finland, its incidence increased during 1990–1999 before showing a sharp downward trend during 2000–2017. Although the Global Report on Diabetes on the WHO website indicates that preventing T1DM is made difficult by its cause being unknown and it not currently being preventable. Investigative efforts have centered on prevention, aiming to either delay or prevent disease onset [
36]. Research from the USA has also shown that the burden of T1DM on the lives of adolescents can be reduced [
37].
The present study found a weak negative association between EAPC and the HDI in 2017, among which EAPC was positively correlated with HDI in type 1 diabetes, and was negatively correlated with HDI in type 2 diabetes. The relationship between EAPC and HDI is different for different types of diabetes, which requires an in-depth analysis of the relationship from a macroeconomic perspective. Association of EAPC and HDI in this study requires further investigation.
This study is subject to several limitations. First, the lack of relevant epidemiological data made it impossible to include some of the important risk factors related to diabetes mellitus. Second, we were not able to model different patterns of certain risk factors, such as different amounts or types of alcohol consumption, the amount of smoking, exercise duration, different BMIs, the birthweight, the socioeconomic status, or public health and medical interventions. While quantifying the risks or causes of these other categories was beyond the scope of this study, we have provided better estimates of the factors influencing diabetes from a global perspective, and this information will be useful when designing new diabetes prevention strategies. Third, the GBD study provides a standardized approach for estimating incidence and prevalence, by geographical regions and countries and aims to use all accessible information on disease occurrence, natural history, and severity that passes a set of inclusion criteria. In this way, there are deviations in estimating different rates for regional and national data, mainly in small countries and less developed countries. In addition, when a (large) number of 95% confidence intervals are calculated, some of them will be expected to fail to contain the underlying quantity.
It is unclear how changing diagnostic criteria would have affected trends in our study. An American study [
38] shows that a doubling of the incidence of diabetes during 1990–2008, and a plateauing between 2008 and 2012, which could be the 1997 change to the diagnostic criteria of diabetes the 1997 change to the diagnostic criteria of diabetes [
39], which lowered FPG from 140 mg/dL or more to 126 mg/dL or more and encouraged a shift from the oral glucose tolerance test to fasting plasma glucose. The global burden of diabetes is enormous and growing, but in some countries it is on the decline or stable. Incidence of diabetes decreased significantly from 2007 to 2014 in Hong Kong Chinese [
40]. Incidence of type 2 diabetes has stabilized in Scotland between 2004 and 2013 [
41].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.