The Incidence of Type 1 Diabetes in the Age Group 0–39 Years Has Not Increased in Antwerp (Belgium) Between 1989 and 2000: Evidence for earlier disease manifestation

The Antwerp district is a geographically defined region that is situated in the northern part of Belgium with a total surface area of 1,000 km² and an average population density of 930 inhabitants per square kilometer (14). During the period 1989–2000, the population size of individuals ages 0–39 years averaged 488,457, based on annual civil registration numbers (14). Those numbers were also used to calculate the male-to-female ratio in the background population and as the denominator for the incidence rates.

During a 12-year study period (1 January 1989 to 31 December 2000), 571 type 1 diabetic patients residing in the Antwerp district and diagnosed before age 40 years were prospectively registered. Registration of newly diagnosed type 1 diabetic patients in the Belgian Diabetes Registry (BDR) is based on voluntary reporting by pediatricians and endocrinologists participating in the BDR, a national data and sample bank for recent-onset diabetic patients and their first-degree relatives under age 40 years. Diabetic patients were diagnosed as type 1 according to National Diabetes Data Group criteria (15). They resided in Antwerp for at least 6 months before clinical onset of diabetes (10). All patients were being treated with insulin; cases secondary to other conditions (e.g., cystic fibrosis, pregnancy, high-dosage steroid treatment) were excluded. Completeness of registration was assessed by the capture-recapture method, which assumes the availability of independent primary and (a combination of) secondary sources (16). By cross-classifying subjects according to their presence or absence in the primary and at least one of the secondary sources, the degree of completeness was estimated according to the method of Bishop et al. (17). Pediatricians and endocrinologists participating in the BDR constituted the primary source of ascertainment; the independent, secondary source of ascertainment was a combination of reporting by general practitioners and diabetes nurses; the Antwerp branch of the Flemish Diabetes Association; and self-reporting by patients (through voluntary participation in a nationwide family study on “Early Diagnosis and Prevention of Diabetes”). The ascertainment rates averaged 96% in the age group 0–14 years and 90% in the age group 15–39 years, and did not change significantly during the observation period.

The epidemiological survey in Antwerp was conducted within the framework of the EUROpe and DIABetes (EURODIAB) ACE and TIGER studies (ages 0–14 years; coordinators: A. Green, University of Aarhus, Denmark, and G. Soltész, University of Pècs, Hungary) and of the Insulin-dependent Diabetes in young Adults (IDA) study (ages 15–39 years; coordinator: K.O. Kyvik, University of Odense, Denmark). To adjust for differences in age and sex between populations and to ensure mutual comparability, directly standardized rates were calculated for the sexes separately and combined, as in the EURODIAB protocol. The common virtual standard population assumes equal numbers in each of the 5-year age categories for each sex.

We determined that 93% of patients diagnosed under age 15 years and 81% of patients diagnosed between ages 15 and 39 years tested positive for at least one type of diabetes-associated autoantibody (directed against islet cell cytoplasm, insulin, IA-2 protein, and/or glutamate decarboxylase) (18). The antibody assays repeatedly achieved 100% diagnostic sensitivity, specificity, and validity in the proficiency testing of the University of Florida (Gainesville, FL) and Louisiana State University (New Orleans, LA). In the combinatorial islet autoantibody workshop (19), assay sensitivity amounted to 73% for islet cell antibody, 36% for insulin autoantibody (IAA), and 85% for glutamate decarboxylase antibody (GADA); IA-2A antibody was not yet available in our laboratory at the time of the workshop (i.e., 1995).

The study was approved by the ethical committees of the universities participating in the scientific projects of the BDR, and informed consent was obtained from each subject and/or his/her parents in accordance with the Helsinki Declaration.

Differences between groups were assessed by Mann-Whitney U test for continuous variables and by χ² test with Yates’correction or Fisher’s exact test, when appropriate, for discontinuous variables. Poisson regression models were used to investigate trends in incidence rate according to age group (0–4, 5–14, 0–14, and 15–39 years), sex, and calendar year and interactions among these variables. Two-tailed statistical tests were performed using SPSS for Windows 10.0 (SPSS, Chicago, IL) for personal computers or STATA for Poisson regression analysis (STATA, College Station, TX). Results were considered significant at P < 0.05.

Between 1 January 1989 and 31 December 2000, 571 residents of the Antwerp district were reported to have developed type 1 diabetes before age 40 years. Figure 1 shows the mean incidence according to age at diagnosis for male and female patients diagnosed during the entire 12-year observation period. In childhood, the incidence was similar in boys and girls, with a peak value between ages 10 and 15 years; thereafter, the incidence tended to decrease more with age in women than in men, especially after age 20 years. To facilitate comparison of our results with other epidemiological studies, such as the EURODIAB (6), IDA (20) (both sponsored by European Union), and DIAbetes MONDiale (DIAMOND; sponsored by the World Health Organization) studies, patients were classified into two age categories for time of diagnosis: 0–14 and 15–39 years. In male subjects, the incidence of type 1 diabetes was similar in the two age groups and averaged 10.8/100,000 per year (Table 1). In the 0- to 14-year-old age group, the disease incidence in girls was slightly, though not significantly, higher than that in boys. In contrast, for diagnosis after age 15 years, the incidence was significantly lower in women than in men (P < 0.001) (Table 1). Between 1989 and 2000, the age- and sex-standardized incidence rate (95% CI) of the disease averaged 9.9/100,000 per year (CI 9.1–10.8) in the age category 0–39 years between 1989 and 2000. Overall, the disease incidence was similar in male and female subjects under age 40 years (Table 1). Regardless of sex, the incidence rate in the age group 0–14 years was slightly higher than in the older group (P < 0.001) (Table 1). The majority of the patients (342 of 571 or 60%) were diagnosed after age 15 years; in this adult-onset group, the male-to-female ratio (1.63) was strikingly increased (P < 0.001) compared to the male-to-female ratio in the background population (1.04) in the age group 15–39 years and to the value in patients diagnosed under age 15 years (0.92). Altogether, there was a slight male excess for diagnosis under age 40 years (1.29; P = 0.01) (Table 1). The sex- and age-dependent differences in diabetes incidence could not be attributed to differences in completeness of case ascertainment according to sex or age (Table 1).

The overall incidence of type 1 diabetes under age 40 years did not increase according to calendar year in the period 1989–2000 (Fig. 2A). However, comparison of the incidence in the age groups 0–14 (Fig. 2B) and 15–39 years (Fig. 2C) showed an increasing trend in the younger age group, but a decreasing trend in the older. These trends could not be explained by temporal changes in ascertainment levels over the study period (Fig. 2).

To further analyze secular trends in incidence according to age and sex, we performed Poisson regression analysis. There was a significant age-by-sex-interaction (P = 0.002) confirming the male predominance in older age groups (10). There was also a significant trend (P = 0.025) in the incidence of type 1 diabetes in the age group 0–14 years compared with in the age group 15–39 years, showing an overall annual increase of 1.8% (CI −2.0 to 5.7) under age 15 years and an annual decrease of 3.8% (CI 0.7–6.7) above that age (Table 2). In children, the increase tended to be most pronounced under age 5 years, where it averaged 5% (CI −3.7 to 14.5, P = 0.06) (Table 2). Preliminary observations suggested that the differences in incidence rates in patients under and over age 15 were primarily attributable to males subjects. The estimated trend in male subjects ages 0–14 years (risk ratio [RR] 1.05, CI 0.99–1.10) was significantly different from subjects ages 15–39 years (RR 0.96, CI 0.92–1.00, P = 0.013), whereas this trend was not significant in female subjects (age 0–14 years: RR 0.99, CI 0.94–1.05; ages 15–39 years: RR 0.97, CI 0.92–1.02; P = 0.46). However, the trend did not differ significantly between male and female subjects under age 15 years.

Our epidemiological survey in the Antwerp district documented that the global incidence of type 1 diabetes arising before age 40 years has not increased over a 12-year period in that region (1989–2000). In line with international epidemiological studies, such as EURODIAB (6), DIAMOND (8), and IDA (20), we classified our patients according to diagnosis before or after age 15 years. Because the increase in incidence of childhood diabetes has been reported to be most pronounced in children under age 5 years, we also studied this age group. The incidence of childhood-onset diabetes in Antwerp is intermediate in Europe, being three times higher than in Macedonia but four times lower than in Finland (6,10). The Belgian incidence rate is similar to reported values for the Netherlands, Germany, and Luxembourg (6,10). In line with other observations in various countries throughout the world (3–6), we noted a tendency toward increasing incidence of childhood-onset type 1 diabetes, especially under age 5 years, during the 12-year observation period in Antwerp paralleled by a decreasing incidence of adult-onset disease. Although this trend is statistically significant, the overlapping CIs between the age groups considered indicated that the observed differences were borderline.

Although a secular trend toward increasing incidence of childhood type 1 diabetes has been reported throughout Europe in both boys and girls, differences according to age and sex were noted between countries (6). The shift toward earlier clinical manifestation of diabetes in the Antwerp district with near complete ascertainment was also confirmed in a larger representative group recruited nationwide during the same period (n = 2,831 of recent-onset Belgian type 1 diabetic patients) (10), but with incomplete case ascertainment (on average 50%) (Table 3) and tended to be restricted to male subjects in both Antwerp (see RESULTS) and nationwide (Table 3). The fraction of boys, but not of girls, diagnosed nationwide under age 15 years and in particular under age 5 years was significantly higher at the expense of the fraction of subjects diagnosed after age 15 years during the period 1995–2000 as compared to the period 1989–1994 (P < 0.002) (Table 3). There were no changes in age distribution of the background population during the same period, according to National Institute for Statistics data (14). Not surprisingly, there was a temporal trend toward an increasing male-to-female ratio in childhood-onset diabetes contrasting with a decreasing ratio in adult-onset disease (Table 3; P = 0.024). This is in line with results of an epidemiological study on childhood diabetes in Malta, reporting a selective decrease of age at diagnosis in boys (21).

The present study suggests that the increasing incidence of childhood diabetes may be due to an earlier clinical presentation rather than to a global rise in diabetes incidence in all age categories. It is conceivable that changes in early exposure to environmental factors may promote more rapid progression of the subclinical disease process to overt diabetes (22). This evolution is especially worrying because of the higher risk for devastating chronic complications in early-onset diabetes (7). A preliminary report from Sweden (23) also describes a decreasing incidence of diabetes diagnosed between age 15 and 35 years at the expense of an increase in incidence of childhood diabetes, suggesting that our observation is not restricted to Belgium but may have more general relevance. Furthermore, comparison of data from countries with diabetes registration up to at least age 29 years, has indicated that the fraction of adult-onset patients tends to be higher in countries with a low incidence of childhood-onset diabetes such as Romania as compared to high-incidence countries such as Sardinia or Sweden; it is conceivable that the observed large geographical differences in incidence of childhood-onset diabetes may at least in part be related to differences in age at diagnosis rather than to lifetime risk (9–13,16,20,23–25).

The present results also confirm a striking male-to-female excess in adult-onset diabetes, which contrasts with the strong female preponderance in other autoimmune diseases (26) and could relate to sex-specific differences in lifestyle, hormonal status, metabolic demands, or genetic susceptibility (8,10,11,24,25). Sex-dependent differences in incidence of type 1 diabetes have, so far, mostly been studied in patients diagnosed before age 15 years (8). In childhood diabetes, the male-to-female ratio has been reported to vary with the incidence level of the population under study. A slight excess of male patients was more often noted in high incidence countries whereas a slight female excess was observed in low incidence countries. In the present study, the male-to-female ratio approximated 0.9 for patients diagnosed under age 15 years, which agrees well with the observed range throughout Europe (0.7–1.3) (8) and the intermediate incidence rate of childhood diabetes in Belgium (6,10).

The tendency of disease incidence and male-to-female ratio to rise in parallel in young Belgian patients is also consistent with the positive correlation between male-to-female ratio and incidence levels in childhood diabetes noted throughout the world (8), and may suggest that a sex-related environmental factor is preferentially predisposing male subjects to rapid diabetes onset. Future research should confirm this observation in other populations and should try to identify this putative selective accelerator of subclinical diabetes in male subjects in the perspective of developing primary prevention (2). Because puberty and pregnancy are believed to precipitate clinical presentation of type 1 diabetes (22,27), an increased metabolic burden on β-cells could represent such an accelerator. Rapid growth rate, cold climate, acute infection, and psychological stress have all been suggested as risk factors for type 1 diabetes increasing the β-cell workload (28). Moreover, experimental studies have shown a rise in islet autoantigen expression in response to increased β-cell metabolism (29). In line with these findings, obesity has recently been proposed as a risk factor for childhood type 1 diabetes (30). The increasing prevalence of childhood obesity (31) and the reported association between (changes in) BMI and autoimmune markers in prediabetic children (32), in male first-degree relatives of diabetic patients (25), and in male middle-aged glucose-intolerant subjects from the general population (33) support the hypothesis that increased metabolic β-cell activity because of obesity and associated insulin resistance—especially in the case of android visceral obesity—constitutes an important effector of the shift toward earlier clinical presentation of not only type 2 but also of type 1 diabetes (34,35). In this respect, it is interesting to note that throughout Europe there is a large male-to-female excess in overweight individuals, with the highest prevalence being noted in Scandinavia and the U.K. (36). In contrast, a striking female-to-male excess exists in underweight subjects, with the highest frequencies of severely underweight women being noted in Luxembourg and Belgium (36).

In conclusion, this study showed that in Belgium, the increasing incidence of childhood type 1 diabetes, especially under age 5 years, is not attributable to a global increase in disease incidence, but rather to earlier clinical manifestation. The results suggest that an environmental factor may preferentially accelerate the subclinical disease process in diabetes-prone subjects. The possible implication of childhood obesity may create perspectives for prevention.

The present work was supported by the Belgian “Fonds voor Wetenschappelijk Onderzoek” (Grants 3-0113-97, 3-0456-99 and Levenslijn-Diabetes Project 7.0021.96) and by Bayer, Life Scan, Novo Nordisk, Ortho Clinical Diagnostics, and Roche. The central unit of the Belgian Diabetes Registry is supported financially by grants from the Ministries of Public Health of the Flemish and French Community.

We thank the “Vlaamse Diabetes Vereniging” and the following members of the Belgian Diabetes Registry for their invaluable help in recruiting diabetic patients and first-degree relatives and/or in the handling of blood samples: E. Anckaert, P. Arnouts, J-P. Baeyens, E. Balasse, D. Beckers, H. Becq, J. Beirinckx, L. Berghmans, W. Bettens, A. Bocquet, A. Bodson, T. Boereboom, R. Bouillon, J. Bourguignon, M. Buysschaert, A. Carlier, Cardon, A. Chachati, S. Claes, I. Claeys, L. Claeys, Ph. Cochez, M. Coeckelberghs, I.M. Colin, J-L. Coolens, P. Coremans, F. Coucke, W. Coucke, E. Couturier, R. Craen, C. Daubresse, K. Decochez, J. De Cock, P. Decraene, I. De Feyter, R. De Hauwere, J. De Lepeleire, B. Delgrange, C. Delvigne, R. Demaeseneer, P. De Monie, G. De Proft, L. Derdelinckx, P. De Roeck, J. De Schepper, S. De Weer, P. De Winter, L. Dooms, H. Dorchy, E. Duvivier, L. Emsens, C. Ernould, A. Eykens, A. Fassotte, F. Féry, N. Gaham, K. Garmijn, J. Gérard, C. Gillet, I. Gios, P. Goetstouwers, C. Guiot, F. Hay, A. Haemers, S. Haemers, J. Helmer, C. Herbaut, G. Heremans, Immegeers, R. Janssen, P. Jopart, J. Koumans, G. Krzentowski, G. Lamberigts, M-C. Lebrethon, P. Lefèbvre, P. Lemmens, M. Letiexhe, I. Leunckens, T-T. Lim, C. Litvine, K. Logghe, M. Maes, W. Maes, Y. Maus, R. Meulepas, J. Mockel, J. Monballyu, G. Moorkens, W. Musch, D. Nicolaij, F. Nobels, G. Op de Beeck, F. Peiffer, M-C. Pelckmans, H. Penninckx, O. Peters, M. Pieron, L. Pleysier, P. Prims, A. Purnode, G. Raes, C. Righes, C. Roose, B. Roth, R. Rottiers, D. Scarniere, A. Scheen, J. Schets, L. Schillemans, I. Schoemaker, S. Schrans, F. Schuit, J. Schutyser, O. Segers, J-C. Sodoyez, G. Somers, D. Staessen, M-P. Stassen, A. Stroobant, P. Taelman, M. Tersago, L. Terriere, J. Teuwen, G. Thenaers, G. Thiry-Counson, P. Thielemans, U. Timmermans, J. Tits, A. Torfs, K. Van Acker, A. Van Baarle, J. Van Boxem, P. Van Crombrugge, L. Van Damme, L. Van de Mierop, P. Van den Bogaert, E. Vandenbussche, B. Van der Auwera, B. Van der Henst, M. Vanderijst, H. Vanderstappen, W. Van Dessel, C. Van Deun, P. Van de Wouwer, D. Van Doorn, E. Van Fleteren, L. Van Gaal, J. Van Haegenborgh, M. Van Hecke, E. Van Hollebeke, W. Van Horen, S. Van Imschoot, R. Van Landeghem, F. Van Loon, S. Vanneste, D. Van Nimmen C. Van Parijs, H. Van Ravensteyn, P. Van Rooy, P. Van Vaerenbergh, J. Verbeeck, R. Verbiest, C. Vercammen, E. Vercruyssen, K. ver Elst, A. Vertruyen, A. Verhaegen, H. Verhaegen, G. Verhaert, E. Weber, I. Weemaes, L. Wouters, and G. Wyffels.

We are indebted to N. Alaerts, M. Bodson, V. Claessens, N. Christophe, A. Demarré, T. Demesmaeker, L. De Pree, S. Exterbille, T. Ghysels, P. Goubert, C. Groven, A. Ivens, D. Kesler, C. Lauwers, F. Lebleu, J. Michiels, E. Quartier, M-P. Roggemans, G. Schoonjans, H. Thomas, and J. Vertommen for excellent technical assistance. We are also indebted to Dr. C. Patterson (Department of Epidemiology and Public Health, Queen’s University, Belfast, N. Ireland) and Prof. L. Kaufman (Department of Biostatistics, Vrije Universiteit, Brussels, Belgium) for their much valued statistical advice.