In reviewing the data from observational studies, it should be noted that such studies are dependent on the accurate classification of diabetes type, but this is often far from simple. For example, diabetic ketoacidosis may occur at initial presentation in people who are eventually found to have type 2 diabetes (i.e., they have elevated C-peptide, an absence of islet cell or anti-GAD (glutamate decarboxylase) antibodies and can survive with adequate glycemic control without insulin therapy) (1), while the presence of obesity in a child developing autoimmune type 1 diabetes may lead to an erroneous diagnosis of type 2 diabetes (2).
Population-based studies, in which all individuals within a geographical target area undergo blood glucose testing, represent the ideal method of determining the prevalence of diabetes, as they capture all the asymptomatic and undiagnosed cases, which may represent over 50% of the total number of cases within a given population. However, there are very few such studies available of type 2 diabetes in children and adolescents. Even in those that have been conducted, the use of the oral glucose tolerance test (OGTT) is uncommon, and the studies often lack the key data required to precisely differentiate type 2 diabetes from type 1 diabetes.
In the USA, the National Health and Nutrition Examination Survey (NHANES) III provided prevalence data (based on self-report and fasting glucose) on a national sample of 2867 adolescents aged 12–19 yr, collected between 1988 and 1994 (3). Thirteen adolescents were found to have diabetes, of whom four were considered to have type 2 diabetes, and all four were either non-Hispanic black or Mexican-American. Although the prevalence of type 2 diabetes is not given in the published reports, it can be calculated to be 0.13%. Unfortunately, differentiation between the types of diabetes was based only on the use of insulin, and so it is uncertain to what extent misclassification occurred. The NHANES data from 1999 to 2002 provide a much more recent update for the US population (4). The prevalence of diagnosed diabetes (types 1 and 2) was 0.5%, with 8 of 18 of these diabetes cases being classified as having type 2 diabetes (on the basis of not using insulin), and a further two cases, being on both insulin and tablets, and hence as probably having type 2 diabetes. Data from a single US school district (5) showed a diabetes prevalence of 0.4%, while a study surveying Mexican-American fourth graders found an overall type 2 diabetes prevalence of 0.3% (6), with both these studies using the OGTT for diagnosis. A study of eighth graders from four schools in the southern US, selected because of the high proportion of children from high-risk ethnic groups (56% Hispanic, mean age 13.6 yr), reported that 6.2% had impaired fasting glucose (IFG) (fasting plasma glucose (FPG) ≥ 6.1 mmol/L), 2.3% had impaired glucose tolerance (IGT), and 0.4% had undiagnosed diabetes (FPG ≥ 7.0 mmol/L) (7). Accurate surveillance of the Pima Indian population with regular OGTTs over the last 40 yr has shown rising rates of glucose intolerance (8). From 1967–1976 to 1987–1996, the prevalence of type 2 diabetes in youth increased from 2.4% in males and 2.7% in females to 3.8% in males and 5.3% in females, the highest rates reported in children to date. A study of American-Indian (AI) and Alaskan Native adolescents, using clinical data from across the Indian Health Service, reported that the prevalence of type 2 diabetes increased by 68% from 1990 to 1998 among those aged 15–19 yr (0.32 to 0.54%) (9).
SEARCH for Diabetes in Youth is a six-site study of population-based, physician-diagnosed diabetes in youth aged <20 yr in the USA and used clinical records, C-peptide levels, and autoantibodies to classify the type of diabetes. Among the 1349 cases of diabetes in children aged 0–9 yr, only 11 were classified as having type 2 diabetes (0.01 per 1000) (10). Among youth aged 10–19 yr, the highest prevalence of type 2 diabetes was observed among AI youth (1.74 per 1000), followed by African-American (AA) youth (1.05 per 1000), Asian/Pacific Islanders (APIs) youth (0.54 per 1000), Hispanic youth (0.48 per 1000), and non-Hispanic white youth (NHW) (0.19 per 1000). In this older age group, type 2 diabetes accounted for 76% of diabetes diagnosed among AI youth, but only 6% in the NHW youth (Fig. 1). More recent data from SEARCH provide information on the annual incidence of type 2 diabetes, which is reported as being very low (0.8 per 100 000) in the 0–9 yr age group, rising to 8.1 per 100 000 and 11.8 per 100 000 in 10–14 yr olds and 15–19 yr olds, respectively (11). Among the 15–19 yr olds, the incidence ranged from 5.6 per 100 000 in NHWs to 17.0–22.7 per 100 000 in Hispanics, AAs, and APIs, and 49.4 per 100 000 in AI youth. Among those aged 0–9 yr, type 1 diabetes accounted for 87–100% of all new cases of diabetes, but in the 10–19 yr old age group, type 2 diabetes accounted for between 15% (NHW) and 86% (AI) of all new cases (Fig. 1). Interestingly, Fig. 1 indicates that the proportion of cases that are due to type 2 diabetes has risen over time in each ethnic group, as all the incident cases (with the higher proportion of type 2 diabetes) occurred in 2000–2003, while the prevalent cases developed between 1982 and 2001. Notably, classification of diabetes type was undertaken in a uniform manner for all cases, and this rise in the proportion attributable to type 2 diabetes has occurred at a time when the incidence of type 1 diabetes is known to have risen.
Figure 1. Percentage of all prevalent and incident cases of diabetes that were classified as type 2 diabetes, according to ethnicity: the SEARCH study (10, 11). AA, African-American; AI, American Indian; API, Asian/Pacific Islander; NHW, non-Hispanic white.
Download figure to PowerPoint
Undoubtedly, the largest surveillance of type 2 diabetes in youth has been undertaken in Japan and Taiwan, where screening programs for renal disease, involving urine dipsticking of millions of schoolchildren, have been harnessed to screen for diabetes, by undertaking an OGTT in those with glycosuria. Significant methodological limitations apply with this design, as glycosuria is a relatively insensitive screening tool for diabetes, picking up only those with more severe hyperglycemia. In the most recent report from the Japanese study, involving almost 9 million children screened between 1974 and 2002 (12), the overall annual incidence of type 2 diabetes was 2.63 per 100 000. The annual incidence after 1981 was significantly higher than that before 1980 (1.73 per 100 000 vs. 2.76 per 100 000; p < 0.0001). However, there was no statistical change in the incidence of type 2 diabetes from 1981 to 2002. A further analysis of this study, extending to 2004 (13), has shown a fall in the incidence of type 2 diabetes in junior high school children, with the incidence declining from a high of approximately 7 per 100 000 between 1979 and 1994 to 3.7 per 1000 000 between 2000 and 2004. The authors relate this improvement to evidence of a fall in the prevalence of obesity, an improvement in diet, and increase in physical activity.
A screening program in which fasting blood glucose was measured in those with persistent glycosuria, carried out in 3 million students (aged 6–18 yr) in Taiwan (14), found the prevalence of undiagnosed diabetes to be 9.0 and 15.3 per 100 000 in boys and girls, respectively. The prevalence of undiagnosed diabetes was 62% higher in girls than boys, after adjustment for other factors, and the cases were most commonly identified between the ages of 12 and 14 yr. A 3-yr follow-up of the clinical outcomes of these cases showed that 54% had type 2 diabetes, 10% had type 1 diabetes, 9% had secondary diabetes, 20% were non-diabetic, and 8% had no definite diagnosis. The cases identified as having type 2 diabetes had a higher mean body mass index (BMI), cholesterol, and blood pressure than those with a normal fasting glucose, suggesting that even at this young age, cardiovascular risk was starting to rise in those with diabetes.
Another very large study recently reported data on over 70 000 17-yr-old Israeli military conscripts (15). Type 2 diabetes (diagnostic fasting, 2 h or random blood glucose, but not treated with insulin) was found in 0.036% of males and 0.01% of females. A study of Turkish adolescents, in which 1647 adolescents had a fasting plasma glucose, identified 1.96% with IFG, but no case of type 2 diabetes (16). Data from a Saudi Arabian study (17) showed that in the under-14-yr-old age group, type 2 diabetes was present in 0.12% and IGT in a further 0.25%. In those aged 14–29 yr, the prevalence of type 2 diabetes was 0.79%, while the prevalence of IGT was 0.21%. Considering the variations in age groups in the studies reported here, the Saudi data point to a relatively high prevalence, and as such are consistent with data in adults showing that the prevalence of diabetes in the Middle-East is probably higher than among any other populations, apart from Pacific Islanders and indigenous peoples from North America and Australasia (18). A school-based study of 2640 adolescents aged 12–19 yr in south India, using fasting blood glucose, found only one case of diabetes, giving a somewhat surprisingly low prevalence of 0.04%, despite almost one in five having a family history of diabetes (19).
Clinic-based and register studies
A large number of studies have published data collected from diabetes clinics and from diabetes registers. These have focused on providing estimates of the ratio of type 1 to type 2 diabetes, as well as descriptions of the phenotype of type 2 diabetes, and some have attempted to estimate the prevalence or incidence of type 2 diabetes, on the assumption that the cases in the clinic represent all those from within a definable population. While the strength of such studies is that the assignment of diabetes type is usually carried out by pediatricians (although not always in a uniform manner), it is often unclear how representative any particular clinic population is. Furthermore, publication bias almost certainly exists, in which clinics finding high rates of type 2 diabetes are likely to analyze and publish their data, while those with low rates of type 2 diabetes are less likely to do so. Regional and national registries should solve most of these problems. Changes in incidence rates over time also need to be interpreted cautiously. At least two factors that are unrelated to actual changes in the incidence of type 2 diabetes may contribute to apparent secular rises in the incidence of type 2 diabetes. First, assignment of diabetes type has varied over time, and while several years ago, almost all youth presenting with diabetes would be assumed to have type 1 diabetes, this is increasingly recognized to be incorrect, and the need for insulin as well as even the presence of diabetic ketoacidosis are now no longer taken to be diagnostic of type 1 diabetes. Second, there is likely to be an element of increasing awareness and surveillance that leads to better identification of undiagnosed type 2 diabetes.
Several studies in the USA have reported increases in the incidence of type 2 diabetes. Data from the medical records of 735 AA and Latino children with insulin-treated diabetes in Chicago (20), showed that type 2 diabetes (based on a clinical diagnosis of type 2 or the presence of markers of insulin resistance or a BMI at diagnosis >27 kg/m2) incidence rates rose by 9% per year from 1985 to 1994, and that the incidence was higher in AAs than Latinos (15.2 vs. 10.7 per 100 000 per year). Presentation was typically around puberty, and 62% of those with type 2 diabetes were girls. Among 1027 consecutive diabetic patients attending a diabetic clinic in Cincinnati (21), a 10-fold increase in type 2 diabetes (defined on clinical criteria, including being non-ketosis prone) incidence rates, from 0.7 per 100 000 per year in 1982 to 7.2 per 100 000 per year in 1994, were observed. Onset was again typically around puberty for type 2 diabetes, and the female : male ratio was 1.7:1. Among 569 children and adolescents presenting to a Florida clinic with diabetes between 1994 and 1998 (22), the proportion with type 2 diabetes (as indicated in medical records) rose from 9.4% of new cases to 20% of new cases over the 5-yr period. Within the group as a whole, being Hispanic, black, or female significantly increased the likelihood of having type 2 diabetes. Similarly, from 1997 to 1999, a study of persons referred to a diabetic clinic in Thailand (23) reported a rise from 5 to 17% in the proportion with type 2 diabetes (as indicated in medical records).
A detailed study from the only pediatric diabetic clinic serving an Australian population of approximately 2 million has documented a rise in the incidence of type 2 diabetes among children aged 0–17 yr (24). Between 1990 and 2002, average annual rises of the incidence of type 2 diabetes (defined by clinical findings, C-peptide level, and an absence of autoantibodies) of 23% in the indigenous population and 31% in the non-indigenous population were observed. By the end of the study period, the estimated annual incidences were approximately 16 per 100 000 in the indigenous population and 1 per 100 000 in the non-indigenous population, which compared to an incidence of just over 20 per 100 000 per year for type 1 diabetes in children nationally. Among the 43 children identified during the study period with type 2 diabetes, the peak age at diagnosis was 13–14 yr, and 65% were girls. In an analysis of 14–20-yr-olds attending an adolescent diabetes clinic in New Zealand, the prevalence of type 2 diabetes (defined as non-ketosis prone and absence of autoantibodies) within the clinic population rose from 1.8% in 1996 to 11.0% in 2002 (25). While only 12.5% of new cases of diabetes were classified as type 2 diabetes in 1997–1999, this figure rose to 35.7% for 2000–2001. Among the 18 patients with type 2 diabetes attending the clinic in 2002, all were either Maori or Pacific Islanders, and in contrast to most other studies, only 50% were female, perhaps reflecting the fact that most of the patients in this study developed diabetes after puberty (mean age of onset 15 yr), while most other studies report populations with a slightly younger diabetes onset, much closer to puberty. Indeed, Taiwanese data reveal that the female bias decreases with increasing age, and within the 16- to 18-yr-old group, the incidence of diabetes was equal in males and females (14).
While there seems to be much evidence of increasing incidence and prevalence of type 2 diabetes among youth in the USA and in other populations, it is possible that this is predominantly a feature of high-risk ethnic groups. A series of studies from Europe indicate that type 2 diabetes remains a rarity in these populations. Well-designed studies from Germany, Austria, France, and the UK (26–28) all show type 2 diabetes accounting for only 1–2% of all cases of diabetes. A survey in which 177 British pediatric diabetes centers reported information on all children (aged 0–16 yr) with diabetes found that less than 1% of all cases were due to type 2 diabetes (defined using the presence of clinical or laboratory evidence of insulin resistance), and the risks of type 2 diabetes were higher in South Asians and in girls (29). In another study from the UK and Ireland (30) between 2004 and 2005, 2665 pediatricians, along with diabetes specialist nurses, provided monthly reports on all new cases of diabetes among children under 17 yr of age. The annual incidence of type 2 diabetes [defined as raised insulin (or C-peptide) and/or the absence of autoantibodies] was estimated to be 0.53 per 100 000, ranging from 0.35 per 100 000 in whites to 1.25 per 100 000 and 3.9 per 100 000 in those of South Asian and black origins, respectively. A single center in France (27) reported that only 2% of 382 children (aged 1–16 yr) with diabetes had type 2 diabetes (diagnosed on the basis of the absence of a need for insulin therapy to ensure survival, absence of autoantibodies, and absence of a human leukocyte antigen DQ genotype associated with a high risk of type 1 diabetes mellitus). Using a national register in Austria, Rami et al. (31) showed that of all newly diagnosed cases of diabetes under the age of 15 in the 3 years from 1999 to 2001, 1.5% of cases were due to type 2 diabetes (defined as overweight at diagnosis, a family history of type 2 diabetes, evidence of insulin resistance, lack of ketonuria at onset, and absence of autoantibodies), giving an incidence of 0.25 per 100 000 per year, which is only one-tenth of the incidence reported in the US SEARCH study (11).
Several possible explanations exist for the much lower numbers of type 2 diabetes reported in these European studies, compared with many other studies. First, this may accurately reflect the differences between populations and may arise from the relatively small numbers of people from high-risk ethnic groups in these European populations. Second, there may be a greater degree of underdiagnosis of type 2 diabetes in parts of Europe than in other parts of the world, although such a difference is not an obvious feature of adult-onset type 2 diabetes. Third, it is noteworthy that most of these European studies draw data from national or regional registers or from a large, national collection of diabetes centers. By comparison, the reports from single centers may be biased to the small number that by chance or because of their particular population, have seen rapid rises in their numbers with type 2 diabetes. However, even the population-based data from the USA (4) showed a prevalence of diagnosed type 2 diabetes among 12- to 19-yr-olds that was approximately 1000-fold higher than the estimates based on the British data (32) on 0- to 16-yr-olds (0.2 vs. 0.0002%). It seems unlikely that such a huge difference exists or can be because of ethnic differences, as even the prevalence in British South Asian children was only 0.001% (32), but the full explanation for these discrepancies remains uncertain.