Global epidemiology of type 1 diabetes in young adults and adults: a systematic review

Background Although type 1 diabetes (T1D) can affect patients of all ages, most epidemiological studies of T1D focus on disease forms with clinical diagnosis during childhood and adolescence. Clinically, adult T1D is difficult to discriminate from certain forms of Type 2 Diabetes (T2D) and from Latent Autoimmune Diabetes in Adults (LADA). We searched the information available worldwide on the incidence of T1D among individuals over 15 years of age, and which diagnostic criteria should be used use to qualify T1D in adults. We then studied the variation of T1D incidence with age in adults, and compared it to the incidence in the <15 years-old. Methods A systematic review of the literature was performed to retrieve original papers in English, French and Spanish published up to November 6, 2014, reporting the incidence of T1D among individuals aged over 15 years. The study was carried out according to the PRISMA recommendations. Results We retrieved information reporting incidence of T1D among individuals aged more than 15 years in 35 countries, and published in 70 articles between 1982 and 2014. Specific anti-beta-cell proteins or C-peptide detection were performed in 14 of 70 articles (20%). The most frequent diagnostic criteria used were clinical symptoms and immediate insulin therapy. Country-to-country variations of incidence in those aged >15 years paralleled those of children in all age groups. T1D incidence was larger in males than in females in 44 of the 54 (81%) studies reporting incidence by sex in people >15 years of age. The overall mean male-to-female ratio in the review was 1.47 (95% CI = 1.33-1.60, SD = 0.49, n = 54, p = <0.0001). Overall, T1D incidence decreased in adulthood, after the age of 14 years. Conclusions Few studies on epidemiology of T1D in adults are available worldwide, as compared to those reporting on children with T1D. The geographical variations of T1D incidence in adults parallel those reported in children. As opposed to what is known in children, the incidence is generally larger in males than in females. There is an unmet need to evaluate the incidence of autoimmune T1D in adults, using specific autoantibody detection, and to better analyze epidemiological specificities – if any – of adult T1D. PROSPERO registration number CRD42012002369. Electronic supplementary material The online version of this article (doi:10.1186/s12889-015-1591-y) contains supplementary material, which is available to authorized users.


Background
The worldwide epidemiology of childhood Type 1 diabetes (T1D) was extensively described in the 6th edition of the International Diabetes Federation (IDF) [1]. Data were retrieved in approximately 45% of the countries [1][2][3][4]. In contrast, we are unaware of a similar review on the worldwide epidemiology of adult T1D diabetes, although T1D is known to occur even late in adults [5][6][7]. A major limitation of the epidemiology of T1D in adults is certainly the difficulty there is to distinguish it from Type 2 diabetes (T2D) requiring insulin treatment or from Latent Autoimmune Diabetes in Adults (LADA), when specific markers of autoimmunity are not searched.
Here, our primary objective was to describethrough a systematic review of the literaturethe available published information on adult T1D incidence, and the diagnostic criteria used for case definition. A secondary objective was to study how the variations of T1D incidence in adults mirrored those in children.

Literature review
A systematic review was conducted according to the PRISMA recommendations to retrieve original papers published in English, French and Spanish up to November 6th, 2014, in peer-reviewed journals reporting the incidence of T1D among individuals aged more than 15 years, in population-based studies (i.e. collected in a defined geographic area [8]) and reporting the diagnostic criteria used to define T1D.
The databases used for the literature search were Medline (PubMed), Google Scholar and Thomson Reuters (Web of Knowledge). The protocol of the search was registered in the International Prospective Register of Systematic Reviews (PROSPERO) and is available on http://www.crd.york.ac.uk/PROSPERO/display_ record.asp?ID=CRD42012002369 (Registration number: 2012:CRD42012002369). Figure 1 presents the flow diagram of the bibliographic search, Additional file 1 for the full electronic search strategy, and Additional file 2 for the PRISMA checklist.

Data collection
For each study, the following information was extracted: the identification of the study: authors, title, journal, publication year, the period and country of study. The country was categorized by its World Health Organization (WHO) region and economic level: high-income (HIGH) or low-and middle-income (LMIC) [9], Articles screened based on title and abstract n = 178 113 Excluded 18 Articles excluded o 6: did not report incidence of T1D o 3: did not give diagnostic criteria o 5: concern LADA o 3: concern individuals aged less than 16  the geographic coverage of the study: nationwide (when the study was performed in the entire nation) and local (when it was restricted to a given region, city, or a geographically defined population), the diagnostic criteria used to define T1D in adults: detection of autoantibodies against beta-cells (such as: islet cell antibody (ICA), insulin autoantibody (IAA), islet antigen-2 autoantibody (IA-2), anti-glutamic acid decarboxylase antibodies (GAD)), measurement of the fasting C-peptide level [7], need for permanent insulin therapy, time when the administration of insulin therapy was started, and clinical signals of T1D diabetes such as ketosis, ketonuria and weight loss, the sources of data/registers reporting T1D incidence in the studies, defined according to LaPorte et al. [10] as: primary source of information: a "well-established system of standardized registries for identifying new cases", for example national or regional registers, secondary source of information: other different sources of cases "that would provide a check on the degree of ascertainment", for example medical records or hospital discharges, and tertiary source of information: a third approach for identifying cases, for example, through surveillance system or death certificates, the reported percentage of completeness/ ascertainment between sources of information reporting incidence [10], the incidence rates reported in the text, tables or graph (expressed as new cases per 100.000 persons/year) by sex and age classes, additional information such as those concerning rural/urban, or ethnic differences.

Data analyses
The country distribution of the T1D incidence information and the analysis of the diagnostic criteria used were performed on the entire set of articles retrieved. For the few papers for which the results were presented by ethnic origin, we estimated the mean value of the incidence for the given period in the countries/regions concerned.

Correlation between adult and children T1D incidences
In the geographical correlation analyses between children and adult incidences, we considered for each country the more recent nationwide study published, or if not available, the last published set of local studies retrieved from a given area in the country; in addition, we included all published papers reporting autoantibodies against beta-cells or C-peptide. To obtain an estimate of the incidence of T1D in children in the countries for which the adult incidence was available, we used the data provided by the same adult paper, when available. The incidence of T1D in children was not available in 9 of these papers included in the geographical correlation analyses. In this case, it was estimated through a separate systematic review focused on the corresponding countries and periods (see Additional file 3).

Statistics
Data were extracted from graphs using GraphClick [11].
The country-to-country co-variation of children and adult incidences was quantified by the Spearman correlation and a linear regression.
The R software (version 3.0.1) was used for statistical and graphic analyses [12].

Description of the information obtained from the systematic review on adult T1D
Seventy articles reporting incidence of T1D in young adults and adults aged over than 15 years concerned one country, and one article concerning two countries were retrieved in this systematic review, resulting in a total of 71 studies covering 35 countries (Table 1). Twenty-four of the 71 studies were nationwide; 43 papers provided information on the T1D incidence in the age class 15-29 years, 26 in the age class 30-59 years, and 6 in the persons aged >60 years.
A primary source of information was reported in 99% (70 of 71) of the studies: among these reported sources, 60% (42 of 70) were from medical/hospital records, 36% (25 of 70) from national or regional registers, and 4% (3 of 70) from other sources, such as communitybased surveys; a secondary source of information was reported in 90% (64 of 71) of the studies: among these reported sources, 58% (37 of 64) were from medical/ hospital records, 16% (10 of 64) from associations of patients, 14% (9 of 64) from drug or supplies prescription registers, 8% (5 of 64) from national or regional registers, and 5% (3 of 64) from death certificates and schools registers; finally, a tertiary source of information was reported in 21% (15 of 71) of the studies: among these reported sources, 27% (4 of 15) were from national or regional registers, 27% (4 of 15) from associations of patients, 20% (3 of 15) from death certificates, 20% (3 of 15) from drug or supplies prescription registers, and 7% (1 of 15) from medical registers; see details in Table 1. Percentage of ascertainment (completeness) between sources of information was evaluated in 53 of 71 (75%) studies. The mean percentage of ascertainment of these 53 studies was 94% (Table 1).
In the group of young adults (15)(16)(17)(18)(19), the lowest incidence of T1D was reported in Mauritius, (1.1/100.000 persons/year) [13], and the highest in Estonia (39.9/ 100.000 persons/year) [19]. In the 70-79 year age group, the lowest incidence was reported in Navarra, Spain WHO Member States are divided into high-income (HIGH) or low-and middle-income (LMIC) states [30]. AA: autoantibodies, NW: Nation-wide study, NA: Unavailable data. (a) When there were diagnostic doubts, (b) Only for patients aged over 40 years at onset, (c) Not performed in all cases; the author of this study was contacted to confirm the proportion of these cases, but by the time of submission of this paper no answer was available. (0.8/100.000 persons/year) [57] and the highest in Kronoberg, Sweden (55/100.000 persons /year) [63]. The details of all retrieved incidence by study and age classes are in Additional file 4: Table S1.

Comparison of adult and children T1D incidences
The variations of incidence of T1D in adults with country and age were studied in each area for which we retrieved information on a geographically defined population. This concerned 35 countries.

Variation of T1D incidence with age in adults
In 23 out of 35 (66%) countries (55 of 71 studies), the incidence of T1D was higher in the age range of 0-14 compared with 15-19 years. When restricted to the 14 reports for which the criteria of diagnosis of T1D were auto-antibodies against beta-cells or C-peptide detection, the variation of adult incidence with age showed a consistent decrease after the age of 14 years (Figure 2 and Additional file 4: Table S1).

Geographical correlation of adult and child T1D incidence
A significant geographical correlation, as measured by the Spearman correlation coefficient, was found between adult T1D incidence and 0-14 incidence in the age clas-  Figure 2 Age variation of incidence from childhood to adult age. On this figure, the adult estimates of incidence were taken from the 14 reports of the systematic review using the autoantibodies/C-peptide as diagnostic criteria. Full lines correspond to articles from which both child as well as adult information could be retrieved. The dotted lines are those for which the child information was searched in the same country as in the adult paper, but was from a different paper (see Additional file 3 for details on this literature search). The corresponding countries are shown as: BE1: Belgium (2007) [30]; BE2: Belgium (2002) [31]; BE3: Belgium (1997) [32]; DK: Denmark [34]; ES1: Spain, Catalonia [54]; ES2: Spain, Navarra (2014) [56]; ES3: Spain, Navarra (2013) [57]; FI: Finland [35]; IR: Iran (Islamic Republic of) [15]; IT: Italy [45,46]; SE: Sweden [63], TW: Taiwan [81]; US: United States of America [74]. and overall in the entire 15-60 group (r = 0.75, p-value: 5.7 × 10 −10 ). The correlation was not significant in the oldest class where sparse data were available, but the relation was similar ( Figure 3).
Comparison of male and female T1D adult incidences T1D incidence was larger in males aged 15 to 39 years than in females in 44 (81%) of the 54 studies reporting incidence by sex (Additional file 5: Table S2). The mean male-to-female ratio in our review was 1.47 (95% CI for mean 1.33-1.60, SD = 0.49, n = 54, p = < 0.0001).

Discussion
A first result of this systematic review is the paucity of data available on adult incidence of T1D as compared to those concerning children. The 71 studies retrieved provided information on adult T1D in only 35 countries, 40% of the 88 countries with primary childhood T1D incidence information in the 6th IDF atlas [1]. A second result is that only a small proportion (n = 14) of the 71 studies used detection of specific autoantibodies and/or dosage of C-peptide [83] as diagnostic criteria of adult T1D.
A third result was that in a majority of the retrieved studies, adult T1D incidence was greater in men than in women, which contrasts with incidence of T1D in children where sex ratio is around one [2,84]. Using comparative data, Karvonen et al. also described a male excess among young adults in the 15-39 years of age [85]. Sex differences in exposure to possible environmental triggers of T1D, in hormonal/genetic susceptibility, in lifestyle have been proposed as possible explanations for this difference [62].
A last striking observation of the current analysis is the strong geographical correlation of the incidences in adults and children. This correlation may be explained by the fact that adults with T1D share the gene alleles known to be associated to incidence of T1D in children, [86,87], and/or some predisposing environmental causes [4]. For example, in a previous study on incidence of T1D in children, a significant positive correlation was detected between the percentage of urban population and the incidence of T1D in children (r = 0.41 p-value: < 0.0001) [4]; in this review a significantly higher urban proportion of T1D incidence among adults was found in 4 of the 7 studies reporting differences between rural vs urban areas [15,21,42,75].
There was an overall decrease of incidence with age in adults and young adults after the age of 14. A second peak of T1D around the age of 50, as described by Krolewski et al. [88], was only reported in 7% (4 of 58) of the studies [18,63,80,89].
The paucity of data made it impossible to document an increase in adult T1D incidence that would parallel the dramatic increase observed in children [2,3,90]. Indeed, successive studies in the same region over different periods reporting incidence in people aged >30 years of age were only found for Belgium [30][31][32], Lithuania [20][21][22] and Sweden [58][59][60][61][62]. Similarly, this review did not dispose of sufficient data to document differences in the clinical presentation of T1D of adults and children as suggested elsewhere [32,40]; indeed only two of the 71 studies describe differences in clinical presentation of T1D between adults and children [89,91].
Improving the quantity and quality of information on adult T1D is not only useful to better understand the epidemiology and natural history of T1D, but can have practical consequences, as delay of T1D diagnosis may mean retardation in insulin treatment, lost opportunities for potential prevention of acute and chronic complications, and even death [92]: in Croatia [18], 14% of the incident cases were identified solely through death certificates, and high mortality was found in the newlydiagnosed T1D aged over 50.

Conclusions
Overall, the results of this systematic review should encourage the launching of epidemiological studies of adult T1D with specific diagnostic criteria.

Availability of supporting data
All the supporting data are included as additional files.

Additional files
Additional file 1: Search equation used for the bibliographic analysis.
Additional file 3: List of selected papers reporting incidence of T1D in 0-14 year-olds in 9 countries.