Published at www.nejm.org October 8, 2008 (10.1056/NEJMoa0804328) |
|
Published at www.nejm.org October 8, 2008 (10.1056/NEJMoa0804328) |
| |||||||||||||||||||||||||||||||||||
Background The 65-kD isoform of glutamic acid decarboxylase (GAD) is a major autoantigen in patients with type 1 diabetes mellitus. This trial assessed the ability of alum-formulated GAD (GAD-alum) to reverse recent-onset type 1 diabetes in patients 10 to 18 years of age.
Methods We randomly assigned 70 patients with type 1 diabetes who had fasting C-peptide levels above 0.1 nmol per liter (0.3 ng per milliliter) and GAD autoantibodies, recruited within 18 months after receiving the diagnosis of diabetes, to receive subcutaneous injections of 20 µg of GAD-alum (35 patients) or placebo (alum alone, 35 patients) on study days 1 and 30. At day 1 and months 3, 9, 15, 21, and 30, patients underwent a mixed-meal tolerance test to stimulate residual insulin secretion (measured as the C-peptide level). The effect of GAD-alum on the immune system was also studied.
Results Insulin secretion gradually decreased in both study groups. The study treatment had no significant effect on change in fasting C-peptide level after 15 months (the primary end point). Fasting C-peptide levels declined from baseline levels significantly less over 30 months in the GAD-alum group than in the placebo group (–0.21 vs. –0.27 nmol per liter [–0.62 vs. –0.81 ng per milliliter], P=0.045), as did stimulated secretion measured as the area under the curve (–0.72 vs. –1.02 nmol per liter per 2 hours [–2.20 vs. –3.08 ng per milliliter per 2 hours], P=0.04). No protective effect was seen in patients treated 6 months or more after receiving the diagnosis. Adverse events appeared to be mild and similar in frequency between the two groups. The GAD-alum treatment induced a GAD-specific immune response.
Conclusions GAD-alum may contribute to the preservation of residual insulin secretion in patients with recent-onset type 1 diabetes, although it did not change the insulin requirement. (ClinicalTrials.gov number, NCT00435981
[ClinicalTrials.gov]
.)
As an alternative to immunosuppression, autoantigens may be used to induce immunologic tolerance.17 Insulin and the 65-kD isoform of glutamic acid decarboxylase (GAD) are major autoantigens in patients with type 1 diabetes18,19 and have been tested in immunomodulation experiments.20 Data from studies of nonobese diabetic (NOD) mice have indicated that GAD prevents type 1 diabetes.21,22 A dose-finding study in patients with latent autoimmune diabetes in adults (LADA) indicated that a primary injection and a booster injection of 20 µg each of recombinant human GAD in a standard vaccine formulation with alum (GAD-alum) might preserve residual insulin secretion without serious adverse effects.23
In the current study, we administered GAD-alum to young patients with type 1 diabetes of recent onset to see whether treatment with the autoantigen would reduce or halt the loss of residual insulin secretion. Here, we report results from the 15-month study period followed by 15 months of further observation.
Methods
This study was approved by the research ethics committee at Linköping University, Linköping, Sweden, and by the regulatory authorities in Sweden. Patients provided written informed assent, and their parents provided written informed consent.
Study Design
The first author designed the study and recruited study sites and patients, oversaw the data analyses, wrote the manuscript, and vouches for the completeness and accuracy of the data and the analyses. The sponsor, Diamyd Medical, provided the study drug, was involved in the design of the study, held the data, performed the statistical analyses, and oversaw the conduct of the trial.
At eight pediatric clinics in Sweden, 118 patients who were between 10 and 18 years of age who had presented with type 1 diabetes within the previous 18 months were screened for the presence of GAD autoantibodies and fasting C-peptide levels above 0.1 nmol per liter (0.3 ng per milliliter). A total of 70 patients were eligible and were randomly assigned to a double-blind treatment with either 20 µg of GAD-alum (Diamyd, Diamyd Medical) (35 patients) or placebo (the same formulation without the recombinant human GAD — i.e., alum alone) (35 patients). The pharmacy at the Malmö University Hospital in Malmö, Sweden, prepackaged identical-appearing ampules containing either GAD-alum or placebo in identical per-patient boxes. The boxes were labeled with treatment numbers, from 1 to 70, according to a computer-generated randomization list produced by Clinical Data Care in Lund, Sweden. A block size of 10 was used, but block size was not revealed until unblinding. There was no stratification. When a study site identified an eligible patient and an appointment for the first injection had been made, the study site ordered a treatment box. The central pharmacy sent treatment boxes in consecutive order as ordered by the study sites. Thus, the treatment was assigned in random order over the total study population and was blinded to all but the pharmacy and the provider of the randomization list during the main study period.
All patients were treated with multiple daily injections of insulin with a target glycated hemoglobin level of less than 6.5%. At baseline, two patients in the placebo group but no patients in the GAD-alum group were using an insulin pump. The trial aimed to evaluate the safety as well as the efficacy of GAD-alum treatment as compared with placebo in preserving residual insulin secretion. The prespecified primary efficacy end point was the change between baseline and month 15 in the fasting C-peptide level, and the prespecified secondary efficacy end points were changes between baseline and various prespecified time points, up to month 30, in fasting and stimulated C-peptide levels and glycated hemoglobin values. Other end points prespecified for formal analysis were insulin requirement, fasting plasma glucose level, fasting C-peptide:plasma glucose ratio, and GAD autoantibody titer.
Each patient received a subcutaneous primary injection of either GAD-alum or placebo on day 1, followed by a boost 1 month later. Patients remained in the clinic for observation for 3 hours after injection.
On day 1 and at months 3, 9, 15, 21, and 30, a 2-hour mixed-meal tolerance test was performed in accordance with a European study on estimation of beta-cell function.24 This test consists of the ingestion, within a 5-minute period, of 6 ml of a liquid meal (Sustacal) per kilogram of body weight (maximum, 360 ml). The meal test was performed in the morning (between 7 and 10 a.m.) after an overnight fast, in which no food or drink (with the exception of water) and no smoking occurred after 10 p.m. the preceding day. The patients took no short-acting insulin for at least 6 hours before the test; however, patients receiving a continuous subcutaneous infusion of insulin continued it at the normal basal rate but received no additional boluses for at least 6 hours before the test.
After completion of the main study period (15 months), the treatment code was revealed to the statistician, the programmer of the SAS statistical package (version 8.2 for Windows), and the sponsor, and the data were analyzed. The analyses included measurements of C-peptide levels, glycated hemoglobin levels, insulin requirement (units per kilogram of body weight and 24-hour requirement), plasma glucose levels, and GAD autoantibody titers.
With unblinding of the data only to those listed in the preceding paragraph, the study continued for an extension period of 15 months. Data through the end of the study, at the month 30 follow-up, are included in the current report. A final analysis after month 30 was prespecified in the protocol.
The T-cell studies reported here were performed under a separate investigator-initiated protocol. Informed consent was obtained for this protocol as described above for the main protocol.
Laboratory Tests
Laboratory analyses were performed at Linköping University in Linköping, Sweden. C-peptide levels were measured in serum samples with the use of a time-resolved fluoroimmunoassay (AutoDELFIA C-peptide kit, Wallac) as the study progressed. Results for each assay were validated with the inclusion of a C-peptide control module containing a high-level control, a medium-level control, and a low-level control (Immulite, Diagnostic Products Corp.). A software program (1224 MultiCalc, Wallac) was used for automatic measurements and calculation of results; measurements were expressed in nanomoles per liter.
The type of HLA-DQ-A1* and -B1* alleles was determined by means of polymerase-chain-reaction (PCR) amplification of exon 2 sequences and hybridization with allele-specific probes detected with the use of a time-resolved fluorescence immunoassay, as previously described.25 As detailed in a population-based Swedish case–control study,26 the patients were then categorized as high risk, moderate risk, or low risk, on the basis of HLA typing.
Glycated hemoglobin was analyzed by an immunologic method and calibrated against the Swedish national standard Mono-S. Levels were continuously checked against the External Quality Assurance in Laboratory Medicine in Sweden reference standard.
Serum GAD autoantibody titers were estimated in duplicate by means of a radiobinding assay using 35S-labeled recombinant human GAD produced through in vitro transcription and translation (with a pEx9 vector). Sepharose protein A was used to separate free GAD from antibody-bound, labeled GAD. Interassay variation was 10% for the negative control and 8% for the positive control. A diabetes autoantibody standardization program (DASP) in which we participated has shown that this assay has a sensitivity of 76 to 80% and specificity of 96 to 99%.
Peripheral-blood mononuclear cells (PBMCs) were isolated from sodium-heparinized venous-blood samples through Ficoll–Paque density-gradient centrifugation (Pharmacia Biotech). One million PBMCs diluted in AIM V medium supplemented with 20 µM β-mercaptoethanol were cultured with GAD antigen (Diamyd Diagnostics) or phytohemagglutinin for purposes of mitogenic stimulation (Sigma) at a final concentration of 5 µg per milliliter at 37°C, in 5% carbon dioxide. Additional wells with medium alone were used as negative controls. After 72 hours, the PBMCs were separated from the supernatant, and both cells and cell supernatant were frozen at –70°C until used for real-time reverse-transcription PCR and multiplex fluorochrome analysis.
The cytokines interleukins 5, 6, 10, 12 (p70), 13, and 17, tumor necrosis factor 
(TNF-), and interferon-
were measured in the cell supernatant from each sample with the use of a Bio-Plex Human Cytokine 8-Plex Panel and a Bio-Plex Cytokine Reagent Kit according to the manufacturer's instructions (Bio-Rad Laboratories). A Luminex 100 instrument (involving xMAP technology, Luminex) was used for quantification of each specific reaction. The median fluorescence intensities were analyzed with the use of STarStation software, version 2.0 (AppliedCytometry). A five-parameter curve fit was applied to each standard. The antigen-induced cytokine levels were calculated by subtracting the spontaneous cytokine secretion for each sample from the response induced by the antigen.
Total RNA was isolated from PBMCs according to the RNeasy 96 spin protocol (Qiagen Sciences) and was quantified through optical densitometry at 260 nm. Using equal amounts of total RNA (7 ng per microliter), we synthesized complementary DNA (cDNA) by using the High Capacity cDNA Archive Kit (Applied Biosystems). Reverse transcription of total RNA to cDNA was performed with the use of the GeneAmp PCR System 2700 (Applied Biosystems). The FAM-labeled primer–probes Hs00203958 and Hs00171257 (Applied Biosystems) were used to estimate transcription levels of the transcription factor forkhead box P3 (FOXP3) and transforming growth factor β (TGF-β), respectively. VIC-labeled primer–probes were used to estimate transcription levels of the endogenous control (18s) ribosomal RNA (Applied Biosystems). Relative transcription was estimated with the use of the comparative cycle-threshold (Ct) method (protocol P/N 4303859B, Applied Biosystems).
Statistical Analysis
Results from a study of latent autoimmune diabetes in adult patients23 suggested that including 35 patients in each study group would provide the study with a statistical power of 80 to 90% for assessing differences in C-peptide levels, with a significance level of 5%, assuming a mean (±SD) difference in fasting C-peptide levels of 0.12±0.15 nmol per liter (0.36±0.45 ng per milliliter) between the two groups. Data management and statistical analysis of clinical data were performed by the contract research organization Trial Form Support in Lund, Sweden, whose services were procured by the sponsor and were overseen by the first author. A prespecified analysis of covariance (ANCOVA) model was used, in which the change from baseline was taken as the response variable, the study treatment as the explanatory variable, and the baseline value as a covariate. Age, sex, duration of diabetes at baseline, GAD autoantibody titer, and HLA type were identified in advance as factors to use in additional exploratory analyses and were used to prespecify subgroups. Descriptive statistics regarding change in fasting and stimulated C-peptide levels for these subgroups was foreseen, but no formal analysis was planned in the protocol.
ANCOVA involves the assumption of normally distributed response data and homogeneity of variances. In our study, given the number of patients and the lack of findings regarding variance heterogeneity, the ANCOVA models used are statistically valid. Missing data were not replaced. For no analysis was there more than one subject with missing data.
In all tests, the null hypothesis was that there was no difference between active treatment and placebo. Two-sided tests were used for all hypotheses, and the P values are presented with 95% confidence intervals. Since there was only one primary analysis, the P values were not adjusted for multiplicity.27 The hypothesis of no difference between the study groups was assessed at months 15 and 30 of follow-up. In the graphs, for completeness, all significant P values are shown. The data management and statistical analysis of the GAD-autoantibody and T-cell data were performed at the University of Linköping in Linköping, Sweden.
Results
Recruitment and Randomization of Patients
Seventy of the 118 patients screened were eligible. Twenty-eight patients screened were negative for GAD autoantibodies, 17 did not meet the C-peptide criterion, 2 failed to meet both of these criteria, and 1 was excluded owing to epilepsy. Screening took place over a 2-week period in January and February 2005. The first injection of GAD-alum or placebo took place in February 2005, and the last patient completed the 30-month visit in October 2007.
All but one patient received two doses of either GAD-alum or placebo (Figure 1). One patient (a girl in the placebo group) was withdrawn from the study after 1 week, owing to confirmed infectious mononucleosis with icterus; she received only one injection. A total of 69 patients, 35 in the GAD-alum group and 34 in the placebo group, were included in the intention-to-treat analysis.
|
Data for the two study groups were similar at baseline, defined as the day of the first injection, before injection (Table 1). The distribution of HLA genotypes did not differ between the GAD-alum group and the placebo group (Table 1). In accordance with accepted procedures, the baseline data were not compared statistically between the two groups.28
|
There were seven serious adverse events in five patients in the GAD-alum group and five serious adverse events in four patients in the placebo group (Table 2). The serious adverse events in the GAD-alum–treated patients were knee trauma, ketoacidosis, lower-limb fracture, and diarrhea, and one patient had ketoacidosis, high glycated hemoglobin, and streptococcal tonsillitis. The serious adverse events in the placebo group were mononucleosis, ankle fracture, cessation of insulin use, and two episodes of hypoglycemia with seizure in one patient. No serious adverse events occurring by month 30 were judged to be related to the study treatment.
|
Prespecified Efficacy End Points
Both study groups showed a progressive decrease from the baseline level in both fasting and stimulated C-peptide secretion, indicating a gradual loss of beta-cell function (Figure 2A and 2B). There was no significant effect of the study treatment on the change in fasting C-peptide level between baseline and month 15 (primary end point). There was a significant effect of the study treatment on the change in fasting C-peptide level by month 30 (P=0.045) (Figure 2A and Table 3) and on the change in the C-peptide:plasma glucose ratio (P=0.02) (Table 3). Stimulated C-peptide secretion, as measured by the area under the curve, decreased significantly less in the GAD-alum group than in the placebo group, both by month 15 (P=0.01) and by month 30 (P=0.04) (Figure 2B and Table 3).
|
|
Exploratory Analyses of Efficacy
The significant effect of the study treatment on the change between baseline and month 30 in fasting and stimulated C-peptide levels remained after adjusting for differences in duration of diabetes, age, sex, and baseline GAD autoantibody levels (data not shown). The subgroups prespecified in the protocol regarding duration of diabetes, age, sex, HLA classification, and baseline GAD autoantibody levels were also investigated for interaction effects (data not shown). Of these, only duration of diabetes had a significant influence on the efficacy of the study treatment (P=0.05 for fasting at month 30 and P=0.03 for stimulated C-peptide level, measured as area under the curve, at months 15 and 30).
An exploratory formal analysis of the prespecified subgroups regarding duration of diabetes showed that, among patients treated less than 6 months after diagnosis, both fasting and stimulated C-peptide secretion, as measured by area under the curve, decreased significantly less in the GAD-alum group than in the placebo group by month 30 (P=0.03 and P=0.04, respectively) (Figure 2C and 2E), whereas no significant difference was observed between the two groups for patients treated 6 months or more after diagnosis (Figure 2D and 2F). The observed treatment effect among patients treated less than 6 months after diagnosis cannot be attributed to outliers (see the figure in the Supplementary Appendix, available with the full text of this article at www.nejm.org).
Effects on the Immune System
In the group treated with GAD-alum, GAD autoantibody levels increased rapidly, reached a maximum at 3 months, and then decreased but remained significantly higher than in the placebo group (Figure 3A).
|
, and TNF-, but not of interleukins 6 and 12, increased in response to in vitro stimulation with GAD in the GAD-alum group between baseline and month 15 (Figure 3B). Fifteen months after the immune intervention, GAD-induced secretion of the cytokines, except interleukin-12, was significantly higher in samples from the GAD-alum group than in samples from the placebo group (Figure 3B). Increased GAD-induced expression of FOXP3 and TGF-β was found at month 15 in cells from the GAD-alum group, as compared with the placebo group (Figure 3C). According to post hoc correlation analyses, the expression of FOXP3 and TGF-β was correlated for the GAD-alum group but not for the placebo group (Figure 3D).
Discussion
Fasting C-peptide levels in patients with recent-onset type 1 diabetes may decrease quite slowly when fasting glucose levels are kept reasonably low.29 Stimulated C-peptide levels have been considered an end point for assessing the preservation of beta-cell function early in the course of the disease, since the levels correlate with both improved glycemic control and fewer microvascular complications.15,16,30 However, we chose fasting C-peptide levels as the primary end point, based on the previous study of GAD-alum use in patients with LADA.23 There was no significant effect of the GAD-alum treatment on the change in fasting C-peptide level after 15 months (the primary end point), though there was an effect on stimulated C-peptide level. After 30 months, both fasting and stimulated C-peptide levels showed a significantly smaller decline in the GAD-alum group than in the placebo group. However, the apparent protective effect of the GAD-alum treatment on C-peptide secretion was seen only in patients treated less than 6 months after diagnosis.
Our results suggest that two injections of 20 µg of GAD-alum may contribute to preservation of residual insulin secretion in patients with recent-onset type 1 diabetes. In our patients with newly diagnosed diabetes, the duration and magnitude of the GAD-alum treatment effect seems to be similar to that reported for anti-CD3 treatment,15,16 but use of GAD-alum does not appear to be accompanied by treatment-related adverse events. Residual insulin secretion, especially in the long term, affects important clinical outcomes.4 This may be explained by improved overall metabolic control, reduced fluctuation in blood glucose levels and, possibly, increased exposure to C-peptide, which itself may have biologic effects.31
During the study, the insulin requirement, plasma glucose, and glycated hemoglobin levels increased in both study groups. As expected, the aggressive treatment to achieve the glycated hemoglobin target resulted in glycated hemoglobin levels in our patients that were similar to those in adolescents in the intensive-treatment group of the Diabetes Control and Complications Trial (DCCT).32
Minor differences in the demographic characteristics of the two study groups are unlikely to be relevant, in our view. GAD autoantibodies in patients with type 1 diabetes are associated with HLA-DQ-A1*0501–DQB1*02.26 However, the distribution of HLA genotypes in the two study groups were similar and cannot explain the differences in C-peptide responses.
The mechanism by which GAD-alum treatment may alter disease progression in type 1 diabetes is not clear. Since the therapy was otherwise similar in the two groups, differences in preserved beta-cell function would not appear to be related to more intense insulin treatment or better metabolic control in the GAD-alum group. Fasting C-peptide level may be influenced by the actual blood glucose level, but we found significant preservation of the C-peptide level both before and after adjustment for blood glucose level.
Studies in NOD mice have shown that GAD can induce a potent regulatory response in mice with established autoimmunity and after the onset of type 1 diabetes.21,22 A previous dose-finding study revealed that GAD-alum induced an increase in the ratio of CD4+CD25+ T cells to CD4+CD25– T cells at 24 weeks, suggesting a possible effect on regulatory CD4+CD25+ T cells.23 In our study, GAD-alum induced an antigen-specific T-cell population with a broad range of cytokines involved in the regulatory process of the immune system. The treatment also induced a long-lasting, specific B-cell memory. Thus, modulation of the general memory immune responses to GAD could be responsible for beta-cell preservation.
In conclusion, treatment with GAD-alum had an effect on slowing the loss of residual beta-cell function up to 30 months after intervention and was associated with GAD-specific immune modulation, although it did not change the insulin requirement. Our results provide preliminary proof of concept; large-scale confirmatory studies with GAD-alum are under way in Europe (NCT00723411 [ClinicalTrials.gov] ) and the United States (NCT00751842 [ClinicalTrials.gov] ).
Supported by grants from Diamyd Medical and the Swedish Child Diabetes Foundation (Barndiabetesfonden).
Dr. Ludvigsson reports receiving grant support from Diamyd Medical for the mechanistic studies and from NovoNordisk for studies on insulin analogues; and Dr. Örtqvist, lecture fees from Medtronic. Mr. Zerhouni reports being an employee of Diamyd Medical and holding stock options in the company. No other potential conflict of interest relevant to this article was reported.
We thank Lena Berglert, Gosia Smolinska, and Ingela Johansson for their skillful laboratory work; Prof. Åke Lernmark for determination of HLA types and classification of HLA risk groups; Profs. Mark Atkinson, Daniel Kaufman, Åke Lernmark, and David Leslie for their valuable input on a draft of the manuscript; Prof. Lernmark for supplying the pEx9 vector; and John Robertson for oversight of the manufacture of the study medication.
Source Information
From Linköping University, Linköping (J.L., M.F., M.H., S.A., M.C., M.P., O.V., R.C.), the Queen Silvia Children's Hospital, Gothenburg (G.F.), Malmö University Hospital, Malmö (S.I.), Regional Hospital Ryhov, Jönköping (C.J.), Southern Älvsborg Hospital, Borås (A.L.), Halmstad County Hospital, Halmstad (N.-Ö.N.), Örebro University Hospital, Örebro (J.Å.), Astrid Lindgrens Children's Hospital, Karolinska University Hospital, Stockholm (E.Ö.), and Diamyd Medical, Stockholm (P.Z.) — all in Sweden; and the National Public Health Institute, Helsinki (O.V.).
This article (10.1056/NEJMoa0804328) was published at www.nejm.org on October 8, 2008. It will appear in the October 30 issue of the Journal.
Address reprint requests to Dr. Ludvigsson at the Division of Pediatrics, Department of Clinical and Experimental Medicine, Faculty of Health Sciences, Linköping University, SE-58185 Linköping, Sweden, or at johnny.ludvigsson{at}lio.se.
References
| |||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
