Low-Dose Anti-Thymocyte Globulin (ATG) Preserves Β-Cell Function

MRIGTEAIS RG N REGIMENS AND DRUGS THERAPIES: EMERGING

Diabetes Care Volume 41, September 2018 1917

Low-Dose Anti-Thymocyte Michael J. Haller,1 Desmond A. Schatz,1 Jay S. Skyler,2 Jeffrey P. Krischer,3 Globulin (ATG) Preserves -Cell Brian N. Bundy,3 Jessica L. Miller,3 b Mark A. Atkinson,1 Dorothy J. Becker,4 Function and Improves HbA in David Baidal,2 Linda A. DiMeglio,5 1c Stephen E. Gitelman,6 Robin Goland,7 New-Onset Type 1 Diabetes Peter A. Gottlieb,8 Kevan C. Herold,9 Jennifer B. Marks,2 Antoinette Moran,10 3 11 Diabetes Care 2018;41:1917–1925 | https://doi.org/10.2337/dc18-0494 Henry Rodriguez, William Russell, Darrell M. Wilson,12 Carla J. Greenbaum,13 and the Type 1 Diabetes TrialNet ATG-GCSF Study Group*

OBJECTIVE A pilot study suggested that combination therapy with low-dose anti-thymocyte globulin (ATG) and pegylated granulocyte colony-stimulating factor (GCSF) preserves C-peptide in established type 1 diabetes (T1D) (duration 4 months to 2 years). We hypothesized that 1) low-dose ATG/GCSF or 2) low-dose ATG alone would slow the decline of b-cell function in patients with new-onset T1D (duration <100 days).

RESEARCH DESIGN AND METHODS A three-arm, randomized, double-masked, placebo-controlled trial was performed 1University of Florida, Gainesville, FL by the Type 1 Diabetes TrialNet Study Group in 89 subjects: 29 subjects randomized 2Diabetes Research Institute, University of to ATG (2.5 mg/kg intravenously) followed by pegylated GCSF (6 mg subcutane- Miami Miller School of Medicine, Miami, FL 3University of South Florida, Tampa, FL ously every 2 weeks for 6 doses), 29 to ATG alone (2.5 mg/kg), and 31 to placebo. 4University of Pittsburgh, Pittsburgh, PA The primary end point was mean area under the curve (AUC) C-peptide during a 5Indiana University, Indianapolis, IN 2-h mixed-meal tolerance test 1 year after initiation of therapy. Significance was 6University of California, San Francisco, San defined as one-sided P value < 0.025. Francisco, CA 7Columbia University, New York, NY 8 RESULTS University of Colorado Barbara Davis Center for Childhood Diabetes, Aurora, CO The 1-year mean AUC C-peptide was significantly higher in subjects treated with 9Yale University, New Haven, CT ATG (0.646 nmol/L) versus placebo (0.406 nmol/L) (P = 0.0003) but not in those 10University of Minnesota, Minneapolis, MN 11 treated with ATG/GCSF (0.528 nmol/L) versus placebo (P = 0.031). HbA was sig- Vanderbilt University, Nashville, TN 1c 12Stanford University, Palo Alto, CA nificantly reduced at 1 year in subjects treated with ATG and ATG/GCSF, P = 0.002 13Benaroya Research Institute, Seattle, WA and 0.011, respectively. Corresponding author: Michael J. Haller, hallemj@ peds.ufl.edu. CONCLUSIONS Received 6 March 2018 and accepted 12 June Low-dose ATG slowed decline of C-peptide and reduced HbA1c in new-onset T1D. 2018. Additionof GCSF did not enhance C-peptide preservation afforded by low-dose ATG. Clinical trial reg. no. NCT02215200, clinicaltrials Futurestudies should be considered to determine whether low-dose ATG aloneor in .gov. combination with other agents may prevent or delay the onset of the disease. This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/suppl/ Type 1 diabetes (T1D) is a T-cell–mediated process characterized by autoimmune doi:10.2337/dc18-0494/-/DC1. destruction of b-cells and a lifelong dependence on exogenous insulin (1). To date, the *A complete list of the members of the Type 1 overwhelming majority of efforts seeking to impede the autoimmune process and Diabetes TrialNet ATG-GCSF Study Group can be reverse hyperglycemia have utilized single immunosuppressive or immunomodula- found in the Supplementary Data online. tory drugs (2–6). While several agents targeting T and B lymphocytes have shown © 2018 by the American Diabetes Association. promise, no single agent has demonstrated long-term success in preserving C-peptide Readers may use this article as long as the work is properly cited, the use is educational and not or reducing HbA1c as a means of standard medical practice (7,8). Combination therapy for profit, and the work is not altered. More infor- has been proposed as a potential strategy toward developing safe and practical mation is available at http://www.diabetesjournals approaches to the preservation of C-peptide in patients with TID (9–11). .org/content/license. 1918 Low-Dose ATG Preserves C-Peptide in T1D Diabetes Care Volume 41, September 2018

Preclinical studies with combination provided written informed consent; par- 2 mg/kg. Most subjects received ATG or therapies were performed using the non- ticipants ,18 years of age signed a study placebo infusion as part of a 2- to 3-day obesediabetic (NOD)mousemodel. Low- assent form. hospitalization, although some subjects dose anti-thymocyte globulin (ATG) in Screening and subsequent study visits were managed entirely as outpatients. combination with agents capable of pro- took place at 14 TrialNet sites in the Premedication for ATG/placebo infu- viding synergy was proposed as a means U.S. (Supplementary Data). We screened sions included oral diphenhydramine of reducing adverse events and improv- 113 patients (aged 12–45 years) diag- 1.25 mg/kg up to 50 mg, oral acetamin- ing efficacy. Notably, granulocyte colony- nosed with T1D for ,100 days. A total of ophen 15 mg/kg up to 650 mg, and stimulating factor (GCSF) was shown to 89 patients were enrolled (from Decem- intravenous methylprednisolone or pla- improve the therapeutic capacity for di- ber 2014 to June 2016) who had at least cebo at 0.25 mg/kg. Subjects who de- abetes reversal in NOD mice when com- one T1D-related autoantibody (micro- veloped serum sickness in the 1–2 weeks bined with low-dose murine ATG (12). insulin autoantibodies [mIAA], tested following ATG infusion were offered oral In fact, combinations including GCSF af- only if duration of insulin therapy was prednisone with dosing and duration of forded greater efficacy in the NOD mouse ,7 days; glutamic acid decarboxylase-65 therapy at the discretion of the site in- than any other single-drug or combina- autoantibodies [GAD-65Ab], islet cell vestigator. Subjects weighing .50 kg tion approach tested. Efforts to use either antigen-512 autoantibodies [ICA-512Ab], were offered 50 mg every 12 h on days GCSF or higher-dose ATG (6.5 mg/kg) as zinc transporter 8 autoantibodies [ZnT8A], 1–3, then 40 mg every 12 h on day 4, 30 mg monotherapy in patients with recent- or islet cell autoantibodies [ICA]) and had every 12 h on day 5, 20 mg every 12 h on onset T1D failed to preserve C-peptide, stimulated C-peptide levels $0.2 nmol/L day 6, and 10 mg every 12 h on day 7. although in post hoc analysis subjects measured during a mixed-meal tolerance Subjects weighing ,50 kg were offered .21 years of age appeared to benefit test (MMTT) conducted at least 21 days 30 mg every 12 h on days 1–3, then 20 mg from 6.5 mg/kg of ATG (13,14). after diagnosis of T1D and within 37 days every 12 h on day 4, 10 mg every 12 h A randomized, placebo-controlled, of randomization. on day 5, and 5 mg every 12 h on days single-masked pilot clinical trial of low- M.J.H. proposed the trial, which was 6 and 7. GCSF or placebo was adminis- dose ATG (2.5 mg/kg) and pegylated conducted under the auspices of the tered every 2 weeks for a total of 6 doses GCSF (6 mg subcutaneously every 2 Type 1 Diabetes TrialNet Study Group. at a dose of 6 mg subcutaneously or, if weeks for 6 doses) was performed in Sanofi (Cambridge, MA) provided Thymo- the patient weighed ,44.5 kg, a dose of patients with established T1D (duration globulin (ATG) but had no involvement 100 mg/kg. All subjects received inten- of T1D 4 months to 2 years). The pilot with study management, data collection, sive diabetes management. The goal was demonstrated that combination ther- data analysis, or manuscript preparation. to achieve intensive glycemic control as apy with low-dose ATG/GCSF preserved Amgen (Thousand Oaks, CA) provided recommended by the American Diabetes C-peptide in established T1D (15,16). Neulasta (GCSF) and placebo for the Association (18). Patients used either Mechanistic examination of blood cells study and similarly had no further involve- multiple daily insulin injections or an in- from treated subjects showed that low- ment. Roche Diabetes Care (Indianapolis, sulin pump. Frequent daily blood glucose dose ATG/GCSF achieved a relative in- IN) provided glucose meters, test strips, monitoring was performed, in some cases crease of regulatory T cells (Treg) in and lancets for diabetes management. with concomitant use of a continuous circulation (17). Notably, the pilot low- glucose monitor. Use of noninsulin phar- dose ATG/GCSF versus placebo study Randomization and Masking maceuticals that affect glycemic control Patients were randomly assigned in a did not include an arm with subjects was not allowed. 1:1:1 ratio, stratified by participating receiving low-dose ATG monotherapy. site, with 29 subjects randomized to re- To explore the potential of low-dose ceive experimental treatment with ATG Laboratory Tests ATG/GCSF or low-dose ATG monother- Blood samples were sent to TrialNet and GCSF, 29 subjects randomized to re- apy to preserve b-cell function in new- core laboratories for central analyses. ceiveATGalone(andGCSFplacebo), onset T1D, the Type 1 Diabetes TrialNet C-peptide levels were measured from and31subjectsrandomizedtoreceive Study Group conducted a three-arm, frozen plasma using a two-site immu- both placebos. Randomization was con- randomized, double-masked, placebo- noenzymometric assay (Tosoh Biosci- ducted centrally at the TrialNet Coordi- controlled trial (low-dose ATG/GCSF, ence, South San Francisco, CA). HbA nating Center. 1c low-dose ATG, and placebo) in persons was measured using ion-exchange high- The study was double-masked. An in- with new-onset T1D (duration of disease performance liquid chromatography (Var- dependent data and safety monitoring ,100 days). iant II, Bio-Rad Diagnostics, Hercules, board reviewed adverse events and CA). Reliability coefficients for each as- study accrual every 6 months and con- RESEARCH DESIGN AND METHODS say were above 0.99 from split dupli- ducted quarterly safety reviews. An in- cate samples. Biochemical autoantibodies Study Design and Patients dependent medical monitor (masked to This study was registered with Clinical- (mIAA, GAD-65Ab, ICA-512Ab, ZnT8A) treatment assignment) reviewed all ac- Trials.gov (NCT02215200) and conformed were measured using radioimmunobind- cruing safety data. to all applicable regulatory requirements. ing assays, and ICA was measured using The protocol and consent documents Procedures indirect immunofluorescence. A routine were approved by appropriate indepen- ATG or placebo was administered at a chemistry panel was performed (Hitachi dent ethics committees or institutional re- dose of 2.5 mg/kg as two divided in- 917 with reagents from Roche Diagnos- view boards. All participants (or parents) travenous infusions of 0.5 mg/kg and tics). Subjects who screened positive for care.diabetesjournals.org Haller and Associates 1919

serum antibodies to hepatitis B surface Prespecified secondary outcomes in- to achieve the required number and the antigen, hepatitis C, or HIV were ex- cluded safety, slope of C-peptide over study was formally closed to accrual. cluded from participation. Human leu- time, and differences in HbA1c,glucose, kocyte antigen class II alleles were and insulin dose over time. Adverse Events Adverse events were defined using Com- measured using PCR amplification and mon Terminology Criteria for Adverse sequence-specific hybridization. CD4 and Sample Size Events (CTCAE) version 4.0 (21). The CD8 cell counts were measured in whole Using standard equations for the com- number of events and number of pa- blood via an FC500 using four-color fluo- parison of two means, a sample size of tients experiencing events were tabu- rescent monoclonal antibody reagents 26 per group with complete data pro- lated. As prespecified, only grades 2 and software for automated analysis vided power of 85% to detect a 50% and higher were counted as adverse (Beckman Coulter, Indianapolis, IN). increase in the geometric-like mean in events. Only subjects who received at either experimental treatment group least one dose of active or placebo Statistical Methods (compared with the control group) test- therapy were included. The Fisher ex- The prespecified primary outcome of this ing at the 0.025 a level (one-sided), act test (22) (one-tailed) was applied to trial was a comparison of the area under assuming the control group geometric- each system pair of 2 3 2tables(each the curve (AUC) of stimulated C-peptide like mean was 0.433 nmol/L and the SD treatment vs. placebo group). Adverse response over the first 2 h of a 4-h (on the transformed scale) was 0.167. event grades were analyzed using the MMTT conducted at the 12-month visit. These estimates were based on previous Wilcoxon rank sum test (22). All anal- End point analyses were based on the TrialNet phase II studies (19). Assuming yses were conducted in either TIBCO protocol prespecified intention-to-treat 10% dropout, the sample size goal was Spotfire S+ 8.2 Workbench or SAS 9.4. cohort, defined as all participants with set at 28 per group. The overall type I measured 1-year C-peptide AUC re- error for the trial was ;0.05 given two RESULTS gardless of treatment compliance. By pairwise tests. Patient Enrollment August 2017, two subjects had with- If both hypotheses were rejected, the A total of 113 subjects were screened for drawn from the study. As such, 87 of plan was to employ a method proposed eligibility, of whom 89 were random- the 89 randomized subjects completed by Simon et al. (20) to select which ized to one of three treatment arms: their 1-year visit MMTT and were included treatment should be considered first ATG/GCSF, ATG only, or placebo (Fig. 1). in the primary outcome assessment. The for further clinical research, with the Clinical and demographic characteris- AUC mean was computed using the trap- understanding that the adverse effect tics were similar between treatment ezoidal rule calculated by the weighted results might influence the choice. groups (Table 1). Compliance with the sum of the timed C-peptide values dur- The design included an adaptive non- study protocol was high (Supplementary ing the MMTT test, then divided by comparative (i.e., the observed interim Tables 1 and 2 and Supplementary Fig. 120 (minutes). treatment effect has no influence on 1A–C). One subject was randomized but The primary statistical hypotheses the re-estimation) procedure for sample withdrew consent prior to receiving any sought to determine whether the ATG size. The SD (specifically, square root of study drug. All remaining participants or ATG/GCSF group AUC C-peptide the residual mean squared error) and received ATG or placebo per protocol. means were greater than the placebo the mean of the control group are two Ninety percent of all subjects received group mean. This analysis was conducted parameters required to determine sam- all 6 doses of GCSF or placebo subcutane- on a transformed scale using the func- ple size goal and could adversely affect ous injections. Of the 511 doses of GCSF tion log(Y + 1) to provide better nor- the statistical power if they differ from cp or placebo that were administered, only mal distributional behavior by the test the initial values used. The plan was to 9 involved reduced doses per protocol: statistic. The comparison of either ex- check these initial values against the 5, 1, and 3 for the ATG/GCSF, ATG only, perimental treatment group with the observed estimates halfway through and placebo groups, respectively. All but placebo group was based on a Wald the trial (;39 patients with primary two subjects had AUC C-peptide assess- test using an ANCOVA model adjusting end point measured). The specified ments completed at the 12-month visit. for sex, baseline age, and baseline procedure was to calculate a weighted As prespecified, these two were not in- C-peptide at the 0.025 a level (one- average of the initial and observed cluded in the analysis. sided). The predicted means and asso- estimates and determine a new sample ciated 95% CI for each treatment group size goal. Due to rapid recruitment, the C-Peptide were determined at the means of the initial target sample size (N = 84) was At 1 year, AUC C-peptide was higher in other covariates. exceeded (N = 89 randomized) in advance subjects treated with low-dose ATG ver- Mean rate of change of C-peptide of half of the cohort reaching the primary sus placebo (P = 0.0003) (Fig. 2A). In those mean AUC was estimated using a mixed- end point. Per protocol, recruitment was receiving ATG/GCSF, the AUC C-peptide effects model with both random inter- paused until the re-estimation procedure was not significantly different from cept and slope adjusting for age, sex, could be performed. The re-estimation placebo (P = 0.031, with significance de- baseline C-peptide AUC mean, and treat- procedure indicated a goal of 27 per fined as P , 0.025 to adjust for mul- ment assignment. group (an increase of 1 from the initial tiple comparisons). At 1 year, the AUC Prespecified subgroup factors included plan). Since the number of patients miss- C-peptide geometric-like means were baseline age, sex, C-peptide, HbA1c,race/ ing their primary end point was much ATG/GCSF 0.528 nmol/L (95% CI 0.435, ethnicity, and human leukocyte antigen. less than expected, accrual was sufficient 0.627), ATG only 0.646 (0.547, 0.750), 1920 Low-Dose ATG Preserves C-Peptide in T1D Diabetes Care Volume 41, September 2018

Figure 1—Consort diagram. A total of 113 subjects were screened for eligibility, of whom 89 were randomized and 87 completed the primary outcome mea- sure at 1 year. One subject was randomized but withdrew consent prior to receiving any study drug. All remaining participants received ATG/placebo infu- sionsasspeciﬁed in the protocol. This included two participants who received reduced doses per protocol speciﬁcations (one ATG and one placebo).

and placebo 0.406 (0.324, 0.494). The lower in subjects treated with ATG/GCSF these patients in order that they could C-peptide responses within treatment and ATG alone than in placebo-treated be included. groups were heterogeneous, as shown subjects (Supplementary Fig. 4). in spaghetti plots of individual AUC Lymphocyte and CD4/CD8 Ratio Patients who received low-dose ATG C-peptide levels (Supplementary Fig. HbA1c At 1 year following therapy, HbA was or low-dose ATG/GCSF experienced re- 2A–C). 1c significantly lower in both experimen- duced total lymphocytes, reduced CD4 A mixed model was employed to fit tally treated groups versus placebo: T cells, and relative preservation of CD8 all the C-peptide measurements through ATG/GCSF versus placebo P = 0.011 and T cells, resulting in a reduction in the 12 months (Fig. 2B). Consistent with the ATG only versus placebo P = 0.002 (Fig. CD4/CD8 ratio in comparison with sub- primary analysis by ANCOVA, the rates 2C). HbA was adjusted for baseline jects who received placebo (Fig. 2E and of decline for both experimental treat- 1c HbA level, age, and sex using ANCOVA. Supplementary Fig. 5). Complete blood ment groups were less than for the pla- 1c counts with differential data are pro- cebo group but not statistically different Insulin Use vided in Supplementary Table 3. from each other. The difference in the Therewerenostatisticallysignificant dif- rate of decline for the ATG/GCSF and ATG ferences in insulin use between either Adverse Events only groups (relative to the placebo group) experimental treatment group or the Adverse events separated by severity was +0.178 and +0.227 nmol/L per year, placebo group (Fig. 2D). Reported insulin (grade) and adverse event category are respectively. AUC C-peptide values ad- use (units/kg/day) was adjusted for base- showninTables2and3,respectively. justed to show equivalent baseline values line insulin use, age, and sex by ANCOVA. Events of grade 2 or higher were counted and change over time in the treatment There were three patients who did not as adverse events per protocol. In all groups are shown in Supplementary Fig. 3. have baseline insulin use reported; the analyses of adverse events, the one pa- AUC glucose during the 2-h MMTT was mean level at baseline was imputed for tient who withdrew prior to receiving care.diabetesjournals.org Haller and Associates 1921

Table 1—Baseline characteristics experimental treatment groups. There ATG and GCSF ATG only Placebo was no difference in AUC C-peptide at Patient characteristics (N = 29) (N = 29) (N = 31) 1 year in subjects who used steroids to treat serum sickness versus those who Age, years Mean 6 SD 17.2 6 5.0 18.1 6 6.9 16.9 6 4.6 were not exposed to steroids. Median 16.4 15.5 15.0 Adverse events labeled as musculo- Range 12.0–32.8 12.4–42.5 12.2–29.3 skeletal/connective tissue system were Male sex, no. of patients (%) 16 (55.2) 17 (58.6) 17 (54.8) significantly higher in those who re- Race, no. of patients (%) ceived GCSF. There were no significant White 28 (96.6) 29 (100.0) 29 (93.5) differences in infection rate, neoplasm, Black 1 (3.4) 0 (0.0) 2 (6.5) or lymphatic cancers when comparing Ethnicity not Hispanic or Latino, ATG/GCSF versus placebo or ATG only no. of patients (%) 28 (96.6) 27 (93.1) 30 (96.8) versus placebo. One patient in the ATG Autoantibodies positive, only group reported a thyroid papillary no. of patients (%) tumor ;1 month following the 2-year GAD65H 23 (79.3) 23 (79.3) 23 (74.2) study end point, and one patient in IA2H 25 (86.2) 23 (79.3) 25 (80.6) ICA* 26 (92.9) 25 (86.2) 22 (71.0) the placebo group reported a benign ZNT8 21 (72.4) 21 (72.4) 22 (71.0) ovarian cyst. Very few severe hypogly- No. of autoantibodies cemic events were reported. Only two positive, no. of patients (%)* patients in the placebo group reported 1 0 (0.0) 3 (10.3) 1 (3.2) a hypoglycemic event, and one patient 2 3 (10.3) 2 (6.9) 3 (9.7) in the ATG/GCSF group reported a hy- 3 2 (6.9) 1 (3.4) 5 (16.1) poglycemic event. 4 11 (37.9) 11 (37.9) 13 (41.9) 5 13 (44.8) 12 (41.4) 9 (29.0) CONCLUSIONS No. of days from diagnosis In this phase IIb clinical trial, treatment to randomization Median 83 81 84 of new-onset T1D (duration of diag- Range 49–97 47–100 52–99 nosis ,100 days) with a single course Weight, kg of low-dose ATG (2.5 mg/kg) preserved Median 62.3 66.4 62.0 C-peptide and reduced HbA1c when Range 39.8–89.1 39.6–92.4 33.8–118 compared with subjects treated with BMI, kg/m2 placebo 1 year after therapy. The com- Median 21.4 22.6 21.8 bination of ATG (2.5 mg/kg) and pegy- Range 16.6–27.7 15.2–32.8 14.3–34.3 lated GCSF (6 mg subcutaneously AUC mean for C-peptide, every 2 weeks for 6 doses) also lowered nmol/L HbA but was not statistically differ- Mean 6 SD 0.793 6 0.321 0.878 6 0.474 0.966 6 0.503 1c Median 0.701 0.757 0.932 ent from placebo with respect to AUC Range 0.33 –1.78 0.211–2.15 0.144–2.08 C-peptide. In contrast to data from NOD mouse HbA1c, %/mmol/mol Median 7.3/56 7.4/57 7.2/55 studies (12), low-dose ATG therapy but Range 5.3–12.3/34–111 4.7–9.0/28–75 5.5–11.2/35–99 not low-dose ATG/GCSF preserved Total daily insulin dose at C-peptide in new-onset T1D. These re- baseline, units/kg* sults suggest that while the NOD model Median 0.339 0.315 0.306 is a useful starting point, species differ- – – – Range 0 1.06 0 0.963 0 0.921 ences (e.g., immune constituents, b-cell *Missing data: one patient missing ICA status (all four other autoantibodies positive); three physiology) may limit the ability of this patients missing baseline insulin dose. animal model to accurately predict effective treatments in human clinical trials. In addition, these data are in contrast any study drug (assigned to ATG/GCSF) steroid treatment in the ATG/GCSF and to the findings in the randomized, single- was removed. No subjects required ex- the ATG only treatment groups, respec- masked pilot study in individuals with tended hospitalization or readmission tively. While the study was formally established T1D, which indicated that due to cytokine release or serum sick- double-masked, adverse reactions may combination therapy with low-dose ATG ness. In addition, no subjects who re- have effectively unmasked subjects as- and GCSF preserves C-peptide (15,16). ceived ATG/GCSF or ATG alone developed signed to active therapy. Presumptions The relatively small sample sizes of the a serious infection. There were no cases of regarding treatment assignment could pilot ATG and GCSF study as well as this grade 4 serum sickness or cytokine release. have led to behaviors that contributed phase IIb effort may have contributed to There were no grade 5 adverse events. to the improved outcomes following the disparate results. Among those having serum sickness, ATG. The use of steroids was not sig- Comparisons of low-dose ATG to pre- 12/20 (60.6%) and 17/21 (81.0%) had nificantly different between the two vious efforts to utilize ATG products are 1922 Low-Dose ATG Preserves C-Peptide in T1D Diabetes Care Volume 41, September 2018

Figure 2—A–E: Effects of low-dose ATG and low-dose ATG/GCSF on C-peptide (A), mixed model predicted C-peptide (B), HbA1c (C), insulin dose (D), and CD4/CD8 ratio (E). A and C–E show adjusted means and 95% CI at each time point. B shows the mixed model predicted population mean of the C-peptide AUC mean by treatment. In all ﬁgures, the placebo group is red, ATG/GCSF is green, and ATG alone is blue.

important. While different from the Sa- patients with new-onset T1D. Additional monotherapy at a dose of 6.5 mg/kg noﬁ ATG product (Thymoglobulin) used studies with ATGAM were abandoned as failed to preserve C-peptide in new-onset in this trial, Eisenbarth et al. (23) re- treatment-associated thrombocytope- T1D subjects aged 12–35 years (24). The ported in 1985 that the combination of nia was thought to outweigh clinical robust cytokine release and serum sick- ATGAM and prednisone reduced HbA1c beneﬁt. In the Study of Thymoglobulin ness associated with higher-dose ATG and induced partial remission in two to Arrest Type 1 Diabetes (START), ATG may in part explain the difference in care.diabetesjournals.org Haller and Associates 1923

Table 2—Adverse events ATG and GCSF ATG only Placebo Adverse effect category Events Patients Events Patients Events Patients Skin and subcutaneous tissue disorder 6 5 (17.9) 10 7 (24.1) 7 4 (12.9) All immune system disorders 33 21 (75.0) 38 23 (79.3) 0 0 (0) Serum sickness only 20 20 (71.4) 21 21 (72.4) 0 0 (0) Cytokine release syndrome only 11 10 (35.7) 17 14 (48.3) 0 0 (0) Musculoskeletal and connective tissue 14 10 (35.7) 3 3 (10.3) 6 5 (16.1) CD4 lymphocyte decrease or other* 42 21 (75.0) 43 22 (75.9) 4 3 (9.7) General disorders and administration** 16 7 (25.0) 18 8 (27.6) 1 1 (3.2) Endocrine disorders 1 1 (3.6) 1 1 (3.4) 7 3 (9.7) Infections and infestations 14 9 (32.1) 9 7 (24.1) 16 9 (29.0) Gastrointestinal disorders 7 5 (17.9) 5 3 (10.3) 6 6 (19.4) Surgical and medical procedures 1 1 (3.6) 0 0 (0) 1 1 (3.2) Psychiatric disorders 7 2 (7.1) 1 1 (3.4) 0 0 (0) Injury, poisoning, and procedural complications 4 1 (3.6) 2 2 (6.9) 5 5 (16.1) Nervous system disorders 4 4 (14.3) 11 4 (13.8) 5 2 (6.5) Metabolism and nutrition disorders 7 4 (14.3) 4 2 (6.9) 4 4 (12.9) Vascular disorders 0 0 (0) 2 2 (6.9) 1 1 (3.2) Neoplasms: benign, malignant, and unspeciﬁed 0 0 (0) 1 1 (3.4) 1 1 (3.2) Respiratory, thoracic, and mediastinal 2 2 (7.1) 3 2 (6.9) 1 1 (3.2) Blood and lymphatic system disorder 1 1 (3.6) 1 1 (3.4) 2 2 (6.5) Cardiac disorders 1 1 (3.6) 0 0 (0) 0 0 (0) Ear and labyrinth disorders 1 1 (3.6) 0 0 (0) 0 0 (0) Total 161 28 (100) 152 29 (100) 67 31 (100) Data are n or n (%). *75% of the events were decreased lymphocytes. Others listed were decreased neutrophils, decreased white blood cells, increased alanine aminotransferase, increased alkaline phosphatase, and increased bilirubin. **Mostly fever and ﬂu-like symptoms.

C-peptide preservation observed with yielding a low CD4/CD8 ratio. The pattern to anti-CD3 (25–27). In contrast, a 50% low-dose ATG (14). of CD4 T-cell reduction with CD8 T-cell reduction in Treg with no change in In this study, we observed the ex- preservation was observed in the pilot CD4/CD8 ratio was observed in START pected reduction in CD4 T cells follow- low-dose ATG/GCSF study (16), corre- (ATG, 6.5 mg/kg). These observations pro- ing treatment with low-dose ATG but latedwithpreservationofTreg,and vide further support for the notion that a relative preservation of CD8 T cells, was associated with clinical responses low-dose ATG and low-dose ATG/GCSF

Table 3—Adverse events grades 3 and 4 ATG and GCSF ATG only Placebo Adverse effect category Events Patients Events Patients Events Patients All immune system disorders 14 12 (42.9) 15 15 (51.7) 0 0 (0) Serum sickness only 12 12 (42.9) 15 15 (51.7) 0 0 (0) Cytokine release syndrome only 2 2 (7.1) 0 0 (0) 0 0 (0) CD4 lymphocyte decrease or other* 24 18 (64.3) 23 17 (58.6) 0 0 (0) General disorders and administration** 4 2 (7.1) 1 1 (3.4) 1 1 (3.2) Endocrine disorders 0 0 (0) 0 0 (0) 5 1 (3.2) Infections and infestations 0 0 (0) 0 0 (0) 1 1 (3.2) Gastrointestinal disorders 1 1 (3.6) 0 0 (0) 0 0 (0) Surgical and medical procedures 0 0 (0) 0 0 (0) 1 1 (3.2) Psychiatric disorders 3 1 (3.6) 0 0 (0) 0 0 (0) Injury, poisoning, and procedural complications 2 1 (3.6) 1 1 (3.4) 1 1 (3.2) Nervous system disorders 1 1 (3.6) 1 1 (3.4) 1 1 (3.2) Metabolism and nutrition disorders 3 2 (7.1) 0 0 (0) 2 2 (6.5) Neoplasms: benign, malignant, and unspeciﬁed 0 0 (0) 1 1 (3.4) 0 0 (0) Blood and lymphatic system disorder 0 0 (0) 1 1 (3.4) 0 0 (0) Total 52 28 43 29 12 31 Data are n or n (%). *Mostly decreased lymphocytes. Others listed were decreased neutrophils, decreased white blood cells, increased alanine aminotransferase, increased alkaline phosphatase, and increased bilirubin. **Mostly fever and ﬂu-like symptoms. 1924 Low-Dose ATG Preserves C-Peptide in T1D Diabetes Care Volume 41, September 2018

are immunologically distinct from higher- compared with placebo. Abatacept, Diabetes TrialNet Study Group is a clinical trials dose ATG. given as 27 intravenous doses over 24 network funded by the National Institutes of Health As we consider the potential application months, resulted in 59% higher ad- (NIH) through the National Institute of Diabetes and Digestive and Kidney Diseases, the National of low-dose ATG to future therapeutic justed C-peptide as compared with pla- Institute of Allergy and Infectious Diseases, and approaches in T1D, we must consider its cebo at the primary outcome period of the Eunice Kennedy Shriver National Institute of limitations. First, the addition of GCSF to 2 years. To compare the relative exper- ChildHealthandHumanDevelopmentthrough low-dose ATG is associated with muscu- imental treatment effects across the the cooperative agreements U01 DK061010, U01 DK061034, U01 DK061042, U01 DK061058, U01 loskeletal side effects but fails to statis- rituximab, abatacept, and ATG/GCSF tri- DK085465, U01 DK085461, UC4 DK085466, U01 ticallypreserveC-peptide.Giventhe costs, als at 1 year, a model was fitted to the DK085476, U01 DK085499, U01 DK085509, U01 challenges of 12 weeks of therapy, and the data of all three studies adjusting for DK097835, U01 DK103266, U01 DK103282, U01 side effects associated with GCSF, there baseline C-peptide, age, and sex. Nota- DK106984, U01 DK107014, and UC4 DK106993; seems little rationale, based on the 1-year bly, low-dose ATG treatment provided the National Center for Research Resources through Clinical Translational Science Awards data, to consider GCSF as a partner for ATG for a 57% increase in AUC C-peptide over UL1TR001085, UL1TR001427, UL1TR001863, in subsequent combination approaches. placebo compared with 16% for rituximab UL1TR001082, UL1TR000114, UL1TR001857, and Second, the use of low-dose ATG does and 23% for abatacept at 1 year. While UL1TR000445; JDRF; and the American Diabetes not fully eliminate side effects associated the modeling accounts for differences Association. Additional funding was provided by with cytokine release or serum sickness. across studies, it should be noted that The Leona M. and Harry B. Helmsley Charitable fi Trust (2015PG-T1D032) and the McJunkin Family That said, the side-effect pro le of low- the rate of C-peptide fall in the ATG study Charitable Foundation. Sanofi provided Thymo- dose ATG, while not inconsequential, is pre- appears to be greater than that in the globulin (ATG), Amgen provided Neulasta (GCSF) dictableandmanageable.Third,giventhat rituximab and abatacept trials. and placebo, and Roche Diabetes Care provided the severity of serum sickness increases Finally, the fusion protein LFA3-Ig diabetes management supplies for the study. with repeated exposure, low-dose ATG is (alefacept) was studied in new-onset Underthetermsof thecooperativeagreement funding mechanism used by the NIH, represen- likely best suited as a classic “induction T1D. Subjects received two 12-week tatives from the sponsoring institutes of the NIH therapy” followed by “consolidation” and courses of the study drug separated by a (National Institute of Diabetes and Digestive and “maintenance” therapies when consider- 12-week pause. While failing to meet its Kidney Diseases) participated in the design and ing future interventions in T1D (28). No- primary end point, a 2-h MMTT (P = 0.065), conduct of the study; interpretation of the data; preparation, review, and approval of the manu- tably, there are ongoing efforts to develop alefacept preserved C-peptide using a script; and the decision to submit the manuscript humanized ATG products that might, if secondary end point, a 4-h MMTT (34). for publication. The sponsor did not have the right effective, markedly reduce if not eliminate Importantly, 42.4% of subjects treated with or ability to veto submission for publication. serum sickness (29). alefacept experienced grade 3 or 4 ad- Duality of Interest. M.A.A. holds patent Low-dose ATG should be compared, verse events. While alefacept remains wor- US8758761B2, Combination therapies for treat- fi ’ ing type 1 diabetes, which includes the use of both in terms of ef cacy and safety, to thy of ongoing study, its manufacturer s ATG in combination with GCSF for the treatment other contemporary agents seeking to voluntary withdrawal of the drug from of type 1 diabetes. Additional funding was pro- preserve C-peptide in T1D. While com- the market challenges its application as vided by Sanofi. No other potential conflicts of parisons to all previous T1D interventions a primary target in future interventional interest relevant to this article were reported. are beyond the scope of this article, we studies. Collectively, the challenges of com- As per the clinical trials agreement between Amgen and NIH and separately between Sanofi compare and contrast several recently paring and contrasting studies with unique and the University of Florida, Amgen and Sanofi studied immunotherapies. The mono- patient populations, end points, and ad- were sent the draft manuscript prior to sub- clonal antibody anti-CD3 has been ex- verse event profiles speaks to the need mission. Amgen and Sanofi did not comment tensively studied in new-onset T1D for multiarm head-to-head comparisons or provide input on the manuscript. Sanofi and (3,27,30,31). Initial studies included sub- of combination T1D immunotherapies. Amgen had no involvement with study management, data collection, data analysis, or manu- jects within 6 weeks of diagnosis who In conclusion, this phase IIb study script preparation. received 14 daily intravenous infusions showed that one course of low-dose Author Contributions. M.J.H., D.A.S., J.S.S., of anti-CD3 hOKT3g1(Ala-Ala). At 1 year, ATG (2.5 mg/kg) within 3 months of and C.J.G. conceived of the study, obtained anti-CD3 treatment provided for signifi- clinical diagnosis of T1D slowed decline funding, collected data, and wrote the manuscript. J.P.K. obtained funding, performed the cant preservation of C-peptide (27). Fe- of b-cell function and reduced HbA1c data analyses, and wrote the manuscript. B.N.B. ver, anemia, and pruritic urticarial rashes for at least 1 year after therapy. Future performed the data analyses and wrote the were common adverse events. Unfortu- studies should include continuous glu- manuscript. J.L.M., M.A.A., D.J.B., D.B., L.A.D., nately, large international studies of anti- cose monitoring to allow assessment of S.E.G., R.G., P.A.G., K.C.H., J.B.M., A.M., H.R., W.R., CD3innew-onsetT1Dfailedtodemonstrate potential effects on glucose variability. and D.M.W. collected data and edited the manuscript. M.J.H. is the guarantor of this work and, C-peptide preservation (30). Long-term follow-up of this study cohort as such, had full access to all the data in the study The Type 1 Diabetes TrialNet has also will determine the duration of protec- and takes responsibility for the integrity of the conducted trials testing anti–B-cell ther- tion afforded by low-dose ATG. Low-dose data and the accuracy of the data analysis. apy with anti-CD20 (rituximab) and co- ATG should be considered, either as a Prior Presentation. Parts of this study were fi stimulation modification with CTLA4-Ig monotherapy or combination therapy, presented at the 78th Scienti c Sessions of the American Diabetes Association, Orlando, FL, (abatacept) in new-onset T1D (32,33). as a potential means for preventing T1D. 22–26 June 2018. Rituximab, administered intravenously in a four-dose course, slowed decline References in C-peptide, reduced HbA1c, and reduced Funding. The sponsor of the trial was the Type 1. Atkinson MA, Eisenbarth GS, Michels AW. insulin dose over a period of 1 year 1 Diabetes TrialNet Study Group. The Type 1 Type 1 diabetes. Lancet 2014;383:69–82 care.diabetesjournals.org Haller and Associates 1925

2. Silverstein J, Maclaren N, Riley W, Spillar R, diabetes in nonobese diabetic mice. Diabetes therapy for patients with recent-onset type 1 Radjenovic D, Johnson S. Immunosuppression 2009;58:2277–2284 diabetes: 2 year results of a randomised trial. with azathioprine and prednisone in recent- 13. Haller MJ, Atkinson MA, Wasserfall CH, et al. Diabetologia 2016;59:1153–1161 onset insulin-dependent diabetes mellitus. N Mobilization without immune depletion fails 25. Tooley JE, Vudattu N, Choi J, et al. Changes in Engl J Med 1988;319:599–604 to restore immunological tolerance or preserve T-cell subsets identify responders to FcR-non- 3. Herold KC, Gitelman SE, Masharani U, et al. A beta cell function in recent onset type 1 diabe- binding anti-CD3 mAb (teplizumab) in patients single course of anti-CD3 monoclonal antibody tes. Clin Exp Immunol. 2016;183:350–357 with type 1 diabetes. Eur J Immunol 2016;46: hOKT3g1(Ala-Ala) results in improvement in 14. Gitelman SE, Gottlieb PA, Rigby MR, et al.; 230–241 C-peptide responses and clinical parameters for START Study Team. Antithymocyte globulin treat- 26. Long SA, Thorpe J, DeBerg HA, et al. Partial at least 2 years after onset of type 1 diabetes. ment for patients withrecent-onsettype1 exhaustion of CD8 T cells and clinical response Diabetes 2005;54:1763–1769 diabetes: 12-month results of a randomised, to teplizumab in new-onset type 1 diabetes. Sci 4. Ludvigsson J, Krisky D, Casas R, et al. GAD65 an- placebo-controlled, phase 2 trial. Lancet Diabe- Immunol 2016;1:eaai7793 tigen therapy in recently diagnosed type 1 di- tes Endocrinol 2013;1:306–316 27. Herold KC, Hagopian W, Auger JA, et al. Anti- abetes mellitus. N Engl J Med 2012;366:433–442 15. Haller MJ, Gitelman SE, Gottlieb PA, et al. CD3 monoclonal antibody in new-onset type 5. Wherrett DK, Bundy B, Becker DJ, et al.; Type Anti-thymocyte globulin/G-CSF treatment pre- 1 diabetes mellitus. N Engl J Med 2002;346: 1 Diabetes TrialNet GAD Study Group. Antigen- serves b cell function in patients with established 1692–1698 based therapy with glutamic acid decarboxylase type 1 diabetes. J Clin Invest 2015;125:448–455 28. Skyler JS. Immune therapy for treating type (GAD) vaccine in patients with recent-onset 16. Haller MJ, Gitelman SE, Gottlieb PA, et al. 1 diabetes: challenging existing paradigms. J type 1 diabetes: a randomised double-blind trial. Antithymocyte globulin plus G-CSF combina- Clin Invest 2015;125:94–96 Lancet 2011;378:319–327 tion therapy leads to sustained immunomod- 29. Kuroiwa Y, Kasinathan P, Sathiyaseelan T, 6. Moran A, Bundy B, Becker DJ, et al.; Type ulatory and metabolic effects in a subset of et al. Antigen-speciﬁc human polyclonal anti- 1 Diabetes TrialNet Canakinumab Study Group; responders with established type 1 diabetes. bodies from hyperimmunized cattle. Nat Bio- AIDA Study Group. Interleukin-1 antagonism in Diabetes 2016;65:3765–3775 technol 2009;27:173–181 type 1 diabetes of recent onset: two multicentre, 17. Couri CE, Oliveira MC, Stracieri AB, et al. 30. Sherry N, Hagopian W, Ludvigsson J, et al.; randomised, double-blind, placebo-controlled C-peptide levels and insulin independence follow- Proteg´ e´ Trial Investigators. Teplizumab for treat- trials. Lancet 2013;381:1905–1915 ing autologous nonmyeloablative hematopoietic ment of type 1 diabetes (Proteg´ e´ study): 1-year 7. Orban T, Bundy B, Becker DJ, et al.; Type stem cell transplantation in newly diagnosed type 1 results from a randomised, placebo-controlled 1 Diabetes TrialNet Abatacept Study Group. diabetes mellitus. JAMA 2009;301:1573–1579 trial. Lancet 2011;378:487–497 Costimulation modulation with abatacept in 18. American Diabetes Association. Glycemic 31. Aronson R, Gottlieb PA, Christiansen JS, patients with recent-onset type 1 diabetes: targets. Sec. 6. In Standards of Medical Care in et al.; DEFEND Investigator Group. Low-dose follow-up 1 year after cessation of treatment. Diabetesd2017. Diabetes Care 2017;40(Suppl. 1): otelixizumab anti-CD3 monoclonal antibody Diabetes Care 2014;37:1069–1075 S48–S56 DEFEND-1 study: results of the randomized 8. Pescovitz MD, Greenbaum CJ, Bundy B, et al.; 19. Bundy BN, Krischer JP; Type 1 Diabetes phase III study in recent-onset human type 1 Type 1 Diabetes TrialNet Anti-CD20 Study TrialNet Study Group. A model-based approach diabetes. Diabetes Care 2014;37:2746–2754 Group. B-lymphocyte depletion with rituximab to sample size estimation in recent onset type 1 32. Pescovitz MD, Greenbaum CJ, Krause- and b-cell function: two-year results. Diabe- diabetes. Diabetes Metab Res Rev 2016;32: Steinrauf H, et al.; Type 1 Diabetes TrialNet Anti- tes Care 2014;37:453–459 827–834 CD20 Study Group. Rituximab, B-lymphocyte 9. Schatz D, Gale EAM, Atkinson MA. Why can’t 20. Simon R, Wittes RE, Ellenberg SS. Random- depletion, and preservation of beta-cell func- we prevent type 1 diabetes? Maybe it’s time ized phase II clinical trials. Cancer Treat Rep 1985; tion. N Engl J Med 2009;361:2143–2152 to try a different combination. Diabetes Care 69:1375–1381 33. Orban T, Bundy B, Becker DJ, et al.; Type 1 2003;26:3326–3328 21. National Cancer Institute. Common Termi- Diabetes TrialNet Abatacept Study Group. 10. Haller MJ, Gottlieb PA, Schatz DA. Type nology Criteria for Adverse Events (CTCAE).Ver- Co-stimulation modulation with abatacept in 1 diabetes intervention trials 2007: where are sion 4.0. U.S. Department of Health and Human patients with recent-onset type 1 diabetes: we and where are we going? Curr Opin Endo- Services, May 2009 (NIH publ. no. 09-5410) a randomised double-masked controlled trial. crinol Diabetes Obes 2007;14:283–287 22. Agresti A. Categorical Data Analysis. New Lancet 2011;378:412–419 11. Ludvigsson J. Combination therapy for pre- York, NY, John Wiley & Sons, 1990 34. Rigby MR, DiMeglio LA, Rendell MS, et al.; servationofbetacellfunctionintype1diabe- 23. Eisenbarth GS, Srikanta S, Jackson R, et al. T1DAL Study Team. Targeting effector memory tes: new attitudes and strategies are needed! Anti-thymocyte globulin and prednisone immu- T cells with alefacept in new-onset type 1 di- Immunol Lett 2014;159:30–35 notherapy of recent onset type 1 diabetes mel- abetes (T1DAL study): 12 month results of a 12. Parker MJ, Xue S, Alexander JJ, et al. Immune litus. Diabetes Res 1985;2:271–276 randomized, double-blind, placeo-controlled depletion with cellular mobilization imparts 24. Gitelman SE, Gottlieb PA, Felner EI, et al.; phase 2 trial. Lancet Diabetes Endocrinol immunoregulation and reverses autoimmune ITN START Study Team. Antithymocyte globulin 2013;1:284–294