Main

During the past decades several facts indicate that autoimmune mechanisms are involved in the etiology of IDDM(1–3). It is clear from studies in identical twins(9), showing a seasonal variation(4–6) and the increasing incidence of IDDM(4, 7, 8), that environmental factors also are involved in the pathogenesis.

In the early 1980s the International Study Group of Diabetes in Children and Adolescents initiated a collaborative prospective study with the aim to follow a risk population consisting of siblings and parents of children with newly diagnosed IDDM to learn more about the etiology and natural history if this disease. We reported elsewhere(10) that patients in Finland and Sweden, countries with a very high incidence of IDDM, seem to have a more aggressive disease process early before (shorter duration of symptoms before diagnosis) and at diagnosis (higher blood glucose values) and a higher frequency of ICA. Although ICA is regarded as secondary to β cell destruction(11, 12), ICSA described by Lernmarket al.(13) in 1978 might play a role in the disease process(12). Even in diseases where a cell-mediated immune reaction seems to be responsible for the main destruction, there is often a humoral component in the early course of an immune response(14, 15). IDDM is nowadays regarded as a T cell-mediated disease, but one could suspect that, along with whatever antigen is involved, B lymphocytes may also react early against that antigen. It has been suggested that, e.g. an antigen with homologous amino acid sequence to certain components in cow's milk protein could be such an antigen(16, 17), and that antigen mimicry could lead to cross-reactivity against the HLA-antigens on the β cell surface. If so, the immune system should in its initial reaction respond with some kind of antibodies toward the cell surface which reflects that the process has been initiated or which might even play a pathogenetic role. Thus, we believe that existence of antibodies toward the cell surface may increase our knowledge of how and when the β cell destruction occurs. In this study we have therefore tried to elucidate the following problems. If an environmental factor elicits the β cell destruction one would expect to find similar environmental effects in several family members simultaneously. Thus, do we find simultaneous ICSA production also in family members? Do ICSA occur more frequently during or just before periods with high incidence of IDDM? Do we more often find ICSA in areas with a high incidence of IDDM?

METHODS

A questionnaire was mailed to those members of the International Study Group of Diabetes in Children and Adolescents who agreed to participate in the study. Nine European clinics participated during certain time periods, and altogether we received background information (year and month of diagnosis, age at diagnosis, sex, duration of symptoms before diagnosis, and so forth) on 718 newly diagnosed children with age at onset < 15 y, and their 2205 relatives, 1346 parents and 859 siblings (Table 1). Background information was given about all their newly diagnosed diabetic patients during such a period, to avoid age, sex, or other selection among those who agreed to let us take blood samples. It has been questioned whether the clinic in Helsinki treats more young or severely ill patients than other hospitals in that geographical area, but the Helsinki patients included in this study had the same age distribution as for diabetic children in the whole of Finland. This indicates that the included patients are representative for the general IDDM population in the area of Helsinki.

Table 1 The participating clinics and number of probands and relatives
Full size table

From the clinics in Sweden, Vienna, and Bologna we have information about the exact number of children acquiring IDDM during the study period in the area of those clinics. The figures indicate that the participating patients do not represent a selected group but the general IDDM population. From the other clinics within the study we do not know the total number of children diagnosed in the different catchment areas. However, those clinics have no selection criteria for admission, and we know that residential area is the dominating reason for people to attend a certain hospital (people go to the nearest hospital). Furthermore, studies from France show that the children in our study have the same age and sex distribution as the children in those studies(18). Thus, in summary we are confident that there is no biased selection of patients referred to or identified within the clinics. We have deliberately divided the clinics into geographical areas. Thus the clinics in Finland and Sweden are called “North.” The clinics in France form another unit, as do the participants in Italy. Central Europe is represented by Berlin and Vienna.

Blood samples were drawn at diagnosis from 576 patients and 1272 first degree relatives, 828 parents and 444 siblings (Table 1). There was no systematic selection bias compared with the total study group. The samples were drawn 1-2 h after breakfast. The serum was divided into 2-mL samples, which were frozen at -20 °C. To the serum sample from which C-peptide later should be analyzed 100 μL of Trasysol per 2 mL of serum were added before freezing.

The C-peptide levels in serum were determined at Novo Research Institute, Copenhagen, according to the method of Heding(19). As these values were nonfasting we regarded values <0.26 pmol/mL as low in our calculations.

ICSA were analyzed in Linköping, Sweden, by using a radioimmunologic method(20). Absorption tests showed that the binding wasβ cell-specific. Interassay variation was 9.5% and intraassay variation 9.9%. For ICSA positivity the specificity was 100% with a sensitivity of 70%(20). Values above mean +2SD for healthy controls were regarded as positive (+), whereas values above mean +1SD was called“weakly positive” (+?). Antibodies against islet cells (ICA and complement fixing-ICA), GPC, TGA, and MCA were determined in London by G. F. Bottazzo(21, 22), whereas blood glucose values were analyzed at each local hospital. MCA and TGA were not determined in about 50% of participants because of lack of serum.

The study has been approved by the Ethical Committee of the Faculty of Health Sciences, Linköping University, Sweden.

STATISTICAL ANALYSIS

Statistical analysis was performed with advice from the Expert group of medical statistics at the University of Linköping by using χ2, correlation matrix (Pearson correlation), multiple stepwise regression, and at test, two-tailed.

RESULTS

Antibodies. In our first study(10) we reported that ICA were more common in patients (74%) than in relatives (5% in parents and 10% in siblings). ICSA were, on the contrary, more often seen in relatives (Table 2), similar to some other autoantibodies. There was no obvious difference between the sexes either in patients or in parents and siblings (Table 2), but there were age differences. Thus, 15% of patients who were less than 5 y old at diagnosis were ICSA-positive and 29% weakly positive compared with only 8 and 14%, respectively, of patients in age group 10-14 y (p < 0.01, in total, and p = 0.07 when including only those strongly positive)(Fig. 1). GPC showed the same pattern as ICSA, whereas TGA and MCA tended to be more common in the oldest age group(Fig. 1). Regarding GPC, boys were responsible for the main difference, whereas an age difference existed in both sexes regarding ICSA. No distinct age differences were seen among the relatives.

Table 2 The frequency of the different antibodies
Full size table
Figure 1
figure 1

The prevalence of different autoantibodies in relation to age at diagnosis.

Full size image

Only regarding ICSA was there a positive correlation between titers in samples from all family members taken at the same occasion (p < 0.01, r = 0.32; multiple stepwise regression). This correlation between family members was seen in all areas but somewhat more pronounced in the North group. Within the families this correlation was strongest between father and patient and second between father and sibling. Fifteen percent of the relatives were positive and 28% weakly positive for ICSA when the proband was positive or weakly positive, compared with only 11% positive and 17% weakly positive relatives when the patient was negative (p < 0.01).

In patients we found a positive correlation between TGA and MCA(r = 0.30, p < 0.01). Also fathers and mothers showed this correlation between TGA and MCA (r = 0.57, p < 0.001 and r = 0.77, p < 0.001, respectively), whereas there was no such correlation between TGA and MCA in siblings.

Geographical differences. In the North group 14% (range 12-16%, NS) of the patients were ICSA-positive compared with only 4% (range 1-6%) in Berlin-Vienna, 10% (range 9-11%, NS) in France, and 4% (range 3-6%) in Italy(North group compared with the other clinics, p < 0.01). If weakly positive ICSA is added, the corresponding figures are 36% (range 33-39%, NS) for North group, 11% (range 9-14%, NS) for Berlin-Vienna, 24%(range 9-20%, NS) for France, and 14% for Italy. North-Berlin-Viennap > 0.01, North-France p < 0.1, North-Italyp < 0.01).

This geographical difference was seen irrespective of age(Fig. 2). GPC showed a similar pattern, although not so pronounced (Fig. 3), whereas MCA and TGA showed no geographical variation.

Figure 2
figure 2

The frequency of ICSA positivity in different age groups and clinics.

Full size image
Figure 3
figure 3

The percentage of ICSA- and GPC-positive patients at onset in the various areas.

Full size image

Regarding ICSA the geographical difference was even more pronounced among relatives; 46% (25%+? and 21%+) in North compared with 21% (range 16-28%,p = 0.1) (17%+? and 4%+) in Berlin-Vienna (p < 0.001), 25% (range 24-28%, NS) (17%+? and 8%+) in France (p < 0.001) and 16% (range 14-29%, Ns) (9%+? and 7%+), in Italy (p < 0.001). We get the same strongly significant differences when comparing only clearly ICSA-positive values. Thus serum from first degree relatives in the nonNordic clinics contain less antibodies binding to the β cell surface than what is seen in the normal healthy control population in Sweden(20). GPC and MCA/TGA on the contrary had the same frequency in relatives all over Europe (Fig. 4).

Figure 4
figure 4

The percentage of ICSA- and CPC-positive relatives at the time of onset in the proband in the various areas.

Full size image

Differences related to some background data. A higher percentage of ICA positivity was seen in relatives of probands with a long duration of symptoms before diagnosis (>60 d) compared with relatives of probands with a short duration of symptoms (<8 d) (13% versus 5%,p = 0.07). In contrast, ICSA (+ and +?) was more common in relatives of patients with a short duration compared with those with a long duration(39% versus 21%, p < 0.05). Corresponding figures for those clearly positive (+) were 15% versus 5% (p < 0.05) (Fig. 5). None of the patients clearly positive for ICSA had a duration of symptoms longer than 60 d before diagnosis, compared with 9% ICSA-negative patients (p = 0.1). Neither ketonuria, serum C-peptide, nor blood glucose at diagnosis showed any relationship to the different antibodies in patients at the time of diagnosis.

Figure 5
figure 5

The frequency of ICA and ICSA positivity in the family members in relation to duration of symptoms before diagnosis (proband).

Full size image

Patients ICSA-positive at diagnosis had more often family members with low postprandial C-peptide (<0.26 pmol/mL) than ICSA-negative patients (14%versus 7%, p < 0.01). There was no significant correlation between ICSA and C-peptide in relatives nor between ICA, GPC, TGA, or MCA and their C-peptide values.

Figure 6 shows the seasonal variation of diagnosis of IDDM (month of first insulin injection) in all patients and the seasonal variation of detectable antibodies (month when the blood sample was taken; most often, the month of the first insulin injection). Among the different antibodies only ICSA seems to have similar seasonality as diagnosis of IDDM, and this seasonal variation is significant (p < 0.05). Also relatives had a tendency to seasonality of ICSA with peaks during spring and autumn.

Figure 6
figure 6

The seasonal variation in the total material (top) and with regard to the different antibodies (bottom). Regarding ICSA, the same results are found, but at a lower level, when including only strongly ICSA-positive patients. TGA and MCA have the same values between months 7 and 12.

Full size image

DISCUSSION

There may be several methodologic problems in a multicenter study of this kind such as selection of patients, varying laboratory methods, and slightly different definitions of the background parameters at the different clinics. However, selection of patients does not seem to explain our results as a great majority of all patients within the catchment areas of the hospitals have been included(10). The clinic in Helsinki was supposed to have more young patients and more patients with severe diabetes compared with other clinics in the area, but fortunately the patients included in this study seem to have the same age distribution as the patients in all of Finland(23). Furthermore, we have compared patients who supplied blood samples with those who did not without finding any differences in age, sex, duration of symptoms before diagnosis, blood glucose values, autoimmune disorder in the family, and so on. We have minimized the risk with varying laboratory methods by using one method at one laboratory for most of the parameters (e.g. C-peptide determined at Novo, Copenhagen, ICSA in Linköping, and the all other antibodies in London).

The samples were frozen, stored, and transported in a similar way. Our results of positivity regarding some antibodies correspond to those of several other studies, both in patients and relatives(24–26). Thus, we believe that in spite of possible methodologic problems there is no reason to expect that they explain our results. Regarding C-peptide we have chosen nonfasting values taken 1-2 h after breakfast to make sure of the best possible compliance from the participating families. It is well known how difficult it is to get fasting values from children, especially difficult when they are healthy children. In any case we are confident in our data because postprandial C-peptide is very well correlated to fasting C-peptide and to 24-h urine C-peptide (J. Ludvigsson, unpublished observations). Furthermore with this method a mean value of 0.45 nmol/L was found in 20 fasting, nondiabetic children, range 0.24-0.72 nmol/L. Faber and Binder(27) found a range of 0.26-0.63 nmol/L in normal subjects after stimulation with glucagon. With this background we found it reasonable to consider a postprandial serum value of<0.26 nmol/L as low.

ICSA were more common in the relatives than in patients. This may have several explanations. Patients who have lost the majority of their β cells have less antigen than the relatives. The process in the patients has proceeded to a later phase when cell-mediated immunity dominates the picture.

In contrast to the other studied autoantibodies, ICSA correlated with the incidence of IDDM in several ways. ICSA were significantly more often found both in patients and in relatives in areas of Europe with the highest incidence of IDDM, and occurrence of ICSA had a seasonal variation very similar to the seasonal variation of IDDM diagnosis. Although there was no obvious relationship between the duration of symptoms before diagnosis and several autoantibodies in the patients, relatives were more often ICSA-positive when the proband had a short duration of symptoms before diagnosis, which is the opposite pattern seen regarding ICA. This also indicates that ICSA is related to onset of IDDM. ICSA-positive patients more often had relatives with low C-peptide values than ICSA-negative patients, which suggests that the β cells in relatives temporarily may be affected(by the same environmental attack?) at the time of the probands diagnosis. Ludvigsson and Afoke(28) and others(29) have suggested that type 1 diabetes can be divided into two groups on the basis of how the disease is elicited and its early manifestations. One group consists of patients diagnosed during a peak seasonal incidence. They tend to have a shorter duration of symptoms, more often ketonuria, lower C-peptide values, and higher HbA1 values than another group of patients diagnosed more evenly throughout the year(28). Our results fit into that pattern as ICSA is most common in both patients with short duration of symptoms before diagnosis and simultaneously in their family members, and ICSA is most common in patients diagnosed just prior and/or during the high incidence season (winter to early spring).

The youngest patients (<5 y at diagnosis) were significantly more often ICSA-positive than older children (>9 y). Perhaps actual environmental factors play a more important role in younger children than in older children in whom increasing insulin-requirements, e.g. during puberty, might cause the manifestation of diabetes. Furthermore, as the clinics in Finland and Sweden have more young children than the other clinics(10), this could imply that environmental factors are more common in the North group, which in turn may explain the lack of ICSA-positive children, >5 y old, in France and Berlin- Vienna.

A common environmental factor seems to cause ICSA production within the whole family as there is a positive correlation between ICSA positivity in the proband and the relatives, also seen in other studies(25, 30). This cannot be explained by multiplex families who were not included in the analysis. All members of certain families tended to produce ICSA simultaneously. The significant correlation between ICSA in fathers and mothers suggests an environmental rather than genetic influence.

Once ICSA were regarded as a significant part of the immune process leading to manifest diabetes(12, 31) and our finding that ICSA-positive patients had more often relatives with low C-peptide values than the ICSA-negative patients supports this idea. Nowadays most facts indicate that the autoimmunity is cell-mediated. ICA is a useful marker reflectingβ cell death and thereby also useful as a marker predicting diabetes. ICSA can probably not be used as such a predictive marker. However, our results suggest that ICSA may be a useful tool to get more information about what is happening during the disease process, and our findings support our earlier results that the geographical difference in incidence of childhood IDDM in Europe parallels geographical differences in autoimmune phenomena both in newly diagnosed diabetic children and their relatives.