Clinical and Immunogenetic Characteristics of Fulminant Type 1 Diabetes Associated with Pregnancy

TYPE 1 DIABETES MELLITUS is characterized by an insulin deficiency resulting from the destruction of pancreatic β-cells (1, 2) and is classified as either type 1A (immune mediated) or type 1B (idiopathic). In type 1A diabetes, islet-related autoantibodies, such as islet cell antibodies, antiglutamic acid decarboxylase (anti-GAD) antibodies (GADab), insulin autoantibodies, and antiislet antigen 2 (IA-2)/IA-2β antibodies (IA-2ab/IA-2βab) are used as markers (3–5). However, these autoantibodies are not always positive in type 1 diabetic patients, even at the onset of overt diabetes. These patients were categorized as having idiopathic (type 1B) diabetes according to American Diabetes Association/World Health Organization criteria. In Japan, fulminant type 1 diabetes, which is included in type 1B diabetes, accounts for approximately 20% of the cases of acute-onset type 1 diabetes and is recognized as a novel subtype of type 1 diabetes (6, 7). The clinical characteristics of this subtype of type 1 diabetes are 1) a remarkably abrupt onset of the disease; 2) very short (<1 wk) duration of diabetic symptoms, e.g. polyuria, thirst, and body weight loss; 3) acidosis at the time of diagnosis; 4) negative status of islet-related autoantibodies, such as islet cell antibodies, GADab, insulin autoantibodies, or IA-2ab; 5) virtually no C peptide secretion (<10 μg/d in the urine); and 6) elevated serum pancreatic enzyme levels (6, 7). Geographic variations exist in the occurrence of type 1 diabetes (8), and one of the lowest age-adjusted incidence rates for the population aged 0–14 yr is found in Japan (9). Little information, however, is available concerning the occurrence of fulminant type 1 diabetes from Western countries (10–12). The incidence of fulminant type 1 diabetes in Caucasians might be relatively small; therefore, this subtype might have been disregarded among the majority of autoimmune type 1 diabetes cases.

Diabetic ketoacidosis (DKA) can be a catastrophic event during pregnancy, complicating approximately 1–3% of diabetic pregnancies (13). Despite intensive therapy, maternal mortality persists, with fetal loss rate reported to be as high as 9–35% (14, 15). One of the characteristics of fulminant type 1 diabetes is that it develops with DKA during pregnancy without preceding diabetes. Shimizu et al. (16) reported that the onset of fulminant type 1 diabetes occurred during pregnancy or after delivery in the Ehime Study. Our previous nationwide survey confirmed that almost all patients who develop type 1 diabetes during pregnancy appeared to have fulminant type 1 diabetes and accounted for 21% of females aged 13–49 yr (7). Autoimmune type 1 diabetes developing during pregnancy is rare in both reports. Therefore, it is important from a clinical point of view to analyze the clinical characteristics of fulminant type 1 diabetes associated with pregnancy (PF).

Susceptibility to type 1 diabetes is determined by both environmental and genetic factors. Although multiple genes have been implicated, human leukocyte antigen (HLA) class II genes, especially the HLA-DR and DQ genes, are the most important and are estimated to account for approximately 50% of the susceptibility to the disease (17, 18). These HLA associations vary depending on geographic and ethnic origins (19). We recently reported that the contribution of HLA class II haplotypes to susceptibility to acute-onset type 1 diabetes depends on the age of onset; namely, the DR9 haplotype is more frequent in adult-onset, but the DR4 haplotype is more frequent in child-onset (20), diabetes. The nation-wide survey of fulminant type 1 diabetes revealed that more than 90% were adult-onset, and only the DR4 haplotype, not DR9, is significantly frequent (21).

It is well known that the immunological milieu is significantly changed during pregnancy. However, the issue of whether the clinical and immunogenetic characteristics are different between the cases associated with and without pregnancy remains to be determined. To investigate the clinical and immunogenetic characteristics of PF patients, we compared them with female patients of child-bearing age with fulminant type 1 diabetes not associated with pregnancy (NPF) in a nationwide survey in Japan.

Patients and Methods

Patients

We selected 161 patients with fulminant type 1 diabetes from the membership of the Japan Diabetes Society, as described previously (7). The criteria used for inclusion of fulminant type 1 diabetes in this study were 1) ketosis or ketoacidosis within 1 wk after the onset of hyperglycemic symptoms; 2) urinary C peptide excretion less than 10 μg/d, fasting serum C peptide levels less than 0.3 ng/ml (0.1 nmol/liter), or serum C peptide levels less than 0.5 ng/ml (0.17 nmol/liter) after glucagon injection or a meal load soon after the onset of the disease; and 3) hemoglobin A_1c levels less than 8.5% on the first visit. These criteria were determined based on data for the first 11 patients with fulminant diabetes reported by Imagawa et al. (6, 7). Among them, 13 patients developed fulminant diabetes with pregnancy. An additional 64 cases were subsequently added to the registration as fulminant type 1 diabetes. Among them, three patients developed fulminant diabetes with pregnancy. Sixteen of the 225 were PF, and an additional six cases have subsequently been added to the list in the principal literature and reports since 2000 until 2004. PF was defined as patients who developed fulminant type 1 diabetes during pregnancy as well as those who developed the disease within 2 wk after delivery. The clinical characteristics of these 22 patients with PF were compared with those of 48 patients with NPF who were selected as female patients of child-bearing age (13–49 yr) from 225 fulminant type 1 diabetes. Of these, the HLA DR and DQ genotypes of the 17 PF and 20 NPF were determined. One hundred ninety nondiabetic subjects without GADab served as controls.

The clinical characteristics of all patients were recorded as described previously. A questionnaire was sent, asking for a description of the patient’s obstetrical state at the onset of fulminant type 1 diabetes, to the following institutes who referred patients to our committee: Ehime University Hospital; Ehime Prefectural Hospital; Fujita Health University School of Medicine; Kagawa Rosai Hospital; Kitakyusyu Municipal Medical Center; Kobe Central General Hospital; Kurasiki Central Hospital; Hiroshima City Asa Hospital; Jichi Medical School; Jichi Medical School, Ohmiya Iryo Center; Nagoya University of Medicine; Niigata City General Hospital; Osaka Medical College Hospital; Showa Medical School; Jikei University School of Medicine, Aoto Hospital; Yokkaichi Municipal Hospital; and Yokohama Municipal Medical Center.

This study was approved by the ethics committee of the Japan Diabetes Society, and informed consent was obtained from all subjects.

Autoantibodies and HLA genotyping

GADab was determined at the time of sampling using a commercially available RIA kit (Cosmic Corp., Tokyo, Japan) as previously described (16). Available DNA samples from 17 PF and 20 NPF patients were genotyped for HLA-DRB1 and HLA-DQB1 as described previously (22).

Statistical analysis

The results reported in this paper are expressed as the mean (range) or the mean ± sd unless otherwise indicated. The χ² test or Fisher’s exact probability test was used to determine the statistical significance of differences between group frequencies. Group comparisons of the levels of clinical parameters were analyzed by the Mann-Whitney’s U test. The prevalence of the DRB1 and DQB1 genotypes in the two clinical phenotype groups was assessed relative to their prevalence in 190 control subjects. The phase or pair of DRB1-DQB1 haplotypes in each subject was inferred manually based on the linkage disequilibrium between the two loci. In the analysis of haplotype combination, increases in prevalence were assessed relative to the prevalence of the neutral genotype combination (X/X; see Table 4), assuming that X/X represents a stable neutral denominator that would be expected to have a similar prevalence in diabetic patients and control subjects (23). The haplotype and genotype frequencies were compared by a χ² test based on a 2 × 2 table.

Results

Clinical and obstetrical characteristics of PF and NPF patients

All patients in the PF and NPF groups had the same characteristics of fulminant type 1 diabetes, as reported previously (Table 1) (7, 16). Among them, arterial pH and hemoglobin A_1c levels at onset of disease in the PF patients were significantly lower than those in the NPF patients (P = 0.0366 and P = 0.0483, respectively). Amylase values in the PF group tended to increase compared with those in the NPF patients (P = 0.0515), but abdominal symptoms were more frequent in NPF compared with PF patients (P = 0.0006). Twenty of 21 PF patients were negative for GADab, and one patient had a transient increase in GADab (12 U/ml). Thirty-eight of 39 NPF patients also were negative for GADab. Eighteen of 22 PF patients developed the disease during pregnancy (7th, 7th, 11th, 17th, 19th, 27th, 29th, 30th, 30th, 30th, 30th, 30th, 31st, 31st, 35th, 36th, 36th, and 38th wk of gestation; average of 18 cases, 26.3 wk), and in 13 of 18 patients, the onset was in the third trimester. These results are consistent with findings for type 1 diabetes reported by Buschard et al. (24). The onset of four PF cases that developed after delivery was within 2 wk after delivery (7th, 10th, 11th, and 14th d; average of 4 cases, 10.5 d). Onset was between spring and summer in more than half of the PF patients. In contrast, the season in which the mother was in her first trimester was winter in more than half of the PF patients.

As shown in Table 2, fetal demise occurred in 12 of 18 PF patients who developed diabetes during pregnancy (67%). The arterial pH in cases of fetal demise was significantly lower than those with live-born infants (P = 0.0053). The duration of hyperglycemic symptoms, such as thirst and polyuria, tended to be shorter in live-born cases (1.8 d; range 1–4 d) than in fetal demise cases (4.2 d; range, 1–9; P = 0.0896). The gestational week at onset was not significantly different between the live-born cases (29 wk; range, 17–36 wk) and the fetal demise cases (25 wk; range, 7–38; P = 0.4323). A cesarean section was performed in five of six live-born cases, whereas this occurred in only three of 12 fetal demise cases (P = 0.0189).

Frequency of HLA DRB1-DQB1 haplotypes in PF and NPF patients

Among the type 1 diabetes-susceptible DRB1-DQB1 haplotypes reported in the Japanese population, DRB1*0901-DQB1*0303(DR9) was found to be significantly more frequent in the PF group than in the NPF group (P = 0.0244) or in control subjects (P = 0.0001), whereas DRB1*0405-DQB1*0401(DR4) was significantly more frequent in the NPF group than in the PF group (P = 0.0162) or in control subjects (P < 0.0001; shown as boldface numbers in Table 3). A high frequency of DR9 in PF was consistent with those with acute-onset type 1 diabetes in adults in the Ehime Study (20), but not for DR4. DRB1*1302-DQB1*0604 was also significantly more frequent in the NPF group than in the control subjects (P = 0.0021), and these data are consistent with findings of acute-onset type 1 diabetes in adult in the Ehime Study (20). However, differences in the frequencies of the other susceptible haplotype, DRB1*0802-DQB1*0302(DR8), and the protective haplotype, DRB1*1502-DQB1*0601 and DRB1*1501-DQB1*0602(DR2), were not significant in either the PF or NPF group compared with the controls.

Effect of DRB1-DQB1 haplotype combination

In the PF group, patients with the DR9 haplotypes tended to have an increased susceptibility to type 1 diabetes in the heterozygous state (not significant), whereas a strong susceptibility was observed when it was present in the homozygote [odds ratio (OR) for DR9/DR9, 10.0; 95% confidence interval (CI), 2.0–49.0; Table 4]. In the NPF group, a similar strong dosage effect of DR4 haplotypes was observed in the homozygote (OR for DR4/DR4, 16.6; 95% CI, 2.6–106.2), whereas the effect of DR4 was weak in the heterozygous state (OR for DR4/X, 4; 95% CI, 1.5–10.8).

Because the OR between heterozygotes and homozygotes of DR4 and DR9 haplotypes differ, the dose effects of these haplotypes (i.e. the number of haplotypes in each subject) were assessed by Armitage’s trend test. In the PF group, the DR9 haplotype showed a significant dosage effect (P < 0.0001), whereas DR4 showed a significant dosage effect in the NPF group (P < 0.0001).

Discussion

One of the lowest incidence rates of childhood-onset type 1 diabetes for the population is found in Japan (9), and adult-onset type 1 diabetes makes up only 2.2% of adult diabetes, as reported in the hospital-based Ehime Study (22). Therefore, it is difficult to collect a large number of type 1 diabetics in our country. The previous nationwide survey revealed that the average incidence of fulminant type 1 diabetes was three cases per month all over Japan, and it accounted for approximately 20% of the cases of acute-onset type 1 diabetes. PF patients accounted for 21% of female fulminant type 1 diabetic patients aged 13–49 yr (7). Therefore, although the patient number is small, the present study represents almost all of the data for PF and NPF patients collected between 2000 and 2004 in Japan.

The present findings suggest that the PF and NPF groups have almost the same phenotype of fulminant type 1 diabetes as those reported in the previous nationwide survey (7). However, some of the clinical findings are different. First, the PF group showed a more severe acidosis than the NPF group at onset, because of accelerated starvation due to hormonal and metabolic changes, an increased incidence of vomiting, and infection during gestation (25). Second, the PF group tended to have higher serum amylase values than the NPF group. An increase in serum exocrine pancreatic enzyme levels is one of the characteristic features of fulminant type 1 diabetes (6, 7). The lower arterial pH as well as the higher amylase values suggest that the PF group may be more clinically severe at the onset of diabetes than the NPF group. Because the onset of classical acute-onset type 1 diabetes during pregnancy is rare in the Japanese (7, 16), we compared the arterial pH of PF patients with those of reported cases of acute-onset type 1 diabetes developed during pregnancy (14, 15). Arterial pH at the onset of disease was significantly lower in the PF patients compared with the reported cases (data not shown), indicating that the lower arterial pH in PF patients may be due to the fulminant form of diabetes as well as the associated pregnancy.

DKA is a medical emergency during pregnancy, not only because of the maternal risk, but also because of high rates of fetal mortality. Fetal demise occurred in 12 of 18 patients (67%) who developed disease during pregnancy. Fetal demise cases showed a more severe acidosis than the live-born cases (P = 0.0053). Montoro et al. (15) reported that fetal loss was 35% in 20 patients with DKA during pregnancy, whereas Cullen et al. (14) reported 9% in 11 patients. Montoro et al. (15) revealed that new unrecognized onset diabetics accounted for 57% of the fetal deaths compared with 21% in mothers with recognized disease. Therefore, the severity of maternal DKA as well as the abrupt onset of diabetes may affect the high fetal mortality rate of PF. The precise mechanism by which maternal DKA affects the fetus is unknown (26, 27). However, some studies suggest a combination of several factors, including maternal dehydration, which diminishes uteroplacental blood flow, or maternal acidosis, leading to fetal acidosis (28). In five of the six live-born cases, a cesarean section rescued the fetuses, whereas in three operated cases of 12 fetal demise patients this was not effective. Because the duration of hyperglycemic symptoms tended to be shorter in the live-born cases (1.8 d; range, 1–4 d) than in the fetal demise cases (4.2 d; range, 1–9 d), it may be possible that fetal lives could be rescued if a cesarean section as well as treatment for DKA were performed immediately after the development of DKA. Thus, all physicians as well as obstetricians should recognize and treat fulminant type 1 diabetes as soon as possible if a pregnant woman abruptly shows hyperglycemic symptoms.

Pregnancy poses an immunological paradox, and maternal T cell tolerance may contribute to fetal survival (29). Wegmann and colleagues (30, 31) proposed that T helper type 2 (Th2) cytokines and helper T cell responses were correlated with successful pregnancy outcomes. Sex hormones, such as progesterone, promote the development of a Th2 response and antagonize the emergence of Th1 cells during pregnancy (32, 33). Type 1A diabetes as well as autoimmune thyroid disease are both considered to be Th1-type diseases. It is well known that autoimmune thyroid disease is ameliorated during pregnancy because of a shift in a Th1- to a Th2-type response, but is aggravated after delivery. This phenomenon is well known to be a postpartum autoimmune disease, especially postpartum thyroid disease (34, 35). Because postpartum aggravation of Hashimoto’s disease usually occurs 1–4 months after delivery, a postpartum rebound in cellular immunity may be assumed to occur around this period (36). However, the onset of disease in the PF group was during pregnancy or immediately after delivery (within 2 wk), and they were negative for GADab, except for a transient increase in one case. Therefore, PF, including cases that developed immediately after delivery, may be caused by a mechanism different from that for postpartum autoimmune thyroid disease. The fact that the onset of classical acute-onset type 1 diabetes is rare during pregnancy (7, 16) suggests that fulminant type 1 diabetes may be caused by a nonautoimmune mechanism, as reported by Imagawa et al. (6)

In contrast, the contribution of the HLA class II haplotype is distinctly different between the PF and NPF groups despite the fact that they are almost the same phenotype. However, it should be noted that these differences may not be conclusive because of the small number of HLA genotypings performed. Viral infection is strongly suggested in fulminant diabetes in both the PF and NPF groups, as reported previously (7, 16) and as evidenced by the frequent flu-like symptoms, and possibly by seasonal variation of onset. Pancreatic islet destruction may be caused by environmental factors, such as viral infections that may trigger the shift from a Th2-type to a Th1-type response, or simply via a bystander effect. The most plausible explanation for the association of the specific HLA class II antigen with autoimmune diseases involves the direct role of the HLA molecule in specific peptide antigen binding and its presentation to CD4⁺ T cells. Different HLA-DR/DQ molecules might have different binding affinities with respect to disease-associated peptides (37). However, the issue of whether the difference between HLA class II molecules between PF and NPF is associated with different disease-associated peptides has not been addressed to date. The human herpes virus-6 (HHV6) virus DNA has recently been detected in a fulminant case associated with the hypersensitivity syndrome (38). High titers of enterovirus IgA antibodies in the serum suggest recurrent enterovirus infections in fulminant type 1 diabetic patients, indicating a higher susceptibility to enteroviral infections (39). Additional work will be needed to determine whether such infections may have pathogenetic importance for the triggering of PF or NPF in different ways or through a common mechanism.

In conclusion, fulminant type 1 diabetes, especially PF, needs to be recognized by all medical staffs because of the abrupt onset during pregnancy or immediately after delivery and the poor prognosis of the resulting fetus. The type 1 diabetes-susceptible HLA class II haplotypes in PF and NPF patients are distinctly different, suggesting that different HLA haplotypes underlie the presentation of PF or NPF.

Acknowledgments

This work was supported by the Japanese Ministry of Health, Labor, and Welfare and a grant from the Japan Diabetes Society.

Abbreviations:

 
• CI,

  Confidence interval;

 
• DKA,

  diabetic ketoacidosis;

 
• GAD,

  glutamic acid decarboxylase;

 
• GADab,

  GAD antibody;

 
• HLA,

  human leukocyte antigen;

 
• IA-2,

  islet antigen 2;

 
• IA-2ab,

  IA-2 antibody;

 
• NPF,

  fulminant type 1 diabetes not associated with pregnancy;

 
• OR,

  odds ratio;

 
• PF,

  fulminant type 1 diabetes associated with pregnancy;

 
• Th2,

  T helper type 2.

References